diff --git a/build/ReleaseManager/osg-release.sh b/build/ReleaseManager/osg-release.sh
index 2d1eda839..7fa28f679 100755
--- a/build/ReleaseManager/osg-release.sh
+++ b/build/ReleaseManager/osg-release.sh
@@ -10,7 +10,7 @@ Buold the source tarball and upload it to the OSG library (now using osg-sw-subm
PRINC - Kerberos principal for the OSG AFS (If only the user is given CS.WISC.EDU is assumed as domain). Krb user may differ form the build machine user name
REPO - url of the Git repository (default: https://github.com/glideinWMS/glideinwms.git)
-h --help - print this message and exit
-
+
NOTE that when using this script you'll have to authenticate with the CILogon URL and insert the OSG PASSWORD (for AFS/Kerberos) and authenticate with the URL again."
[ "$1" == "-h" ] || [ "$1" == "--help" ] && { echo "$USAGE_STR"; exit 0; }
diff --git a/doc/history.html b/doc/history.html
index a33ec55a6..c0e7bb257 100644
--- a/doc/history.html
+++ b/doc/history.html
@@ -134,2240 +134,2979 @@ Stable Series
--
- v3_10_8 released on November 21, 2024 (Manual,
- Installation instructions)
-
- - Advertising information about unprivileged user namespaces in glidein classad (PR#416)
- - Added option --group-name option to manual_glidein_submit
- - Bug fix: Fixed root unable to remove other users jobs in the Factory (PR#433)
- - Bug fix: HTCondor TRUST_DOMAIN configuration macro set to string to avoid Glidein config error (PR#420)
- - Bug fix: Disabled shebang mangling in rpm_build to avoid gwms-python not finding the shell (Issue#436, PR#437)
- - Bug fix: Dynamic creation of HTCondor IDTOKEN password (Issue#440, PR#441)
- - Bug fix: Autodetect CONDOR_OS in the manual_glidein_submit tool (Issue#449, PR#453)
- - Bug fix: Failed log rotation due to wrong file creation time (Issue#451, PR#457)
- - Bug fix: Replacing xmlrunner with unittest-xml-reporting (PR#428)
- - Bug fix: Updated the release upload script to work with osg-sw-submit (PR#439)
-
-
--
- v3_10_7 released on June 21, 2024 (Manual,
- Installation instructions)
-
- - Apptainer cache and temporary directory set in the Glidein working directory (Issue#403, PR#404)
- - Ability to set a minimum required memory for partitionable Glideins. The default is the value used previously, 2500 MB (Issue#405, PR#406)
- - Blackhole Detection. Stop accepting jobs if they are consumed at a rate higher than the configured limit and declare the Glidein a blackhole (Issue#331, PR#399)
- - Bug fix: Fixed Apptainer validation not considering uid_map w/o initial blank (Issue#395, PR#396)
- - Bug fix: Flattening error message in \_CONDOR_WRAPPER_ERROR_FILE and JobWrapperFailure Ad. This is a workaround for a bug in HTCondor (PR#400)
- - Bug fix: Fixed problem when check_signature in glidein_startup is not defined (PR#402)
- - Bug fix: get_tarballs look for HTCondor releases also in the update directory (Issue#412, PR#413)
-
-
--
- v3_10_6 released on January 25, 2024 (Manual,
- Installation instructions)
-
- - Add knobs to allow overloading of memory, GLIDEIN_OVERLOAD_MEMORY, and CPU, GLIDEIN_OVERLOAD_CPUS. (Issue#370, PR#374)
- - Added HTCondor tarball downloader (Issue#367, PR#366)
- - Added default (/bin,/usr/bin) when PATH is empty in glidein_startup.sh (PR#373)
- - Advertising Factory's HTCondor submit parameters (Issue#307, PR#382)
- - Bug fix: Changed M2Crypto imports to be compatible with 0.40.0 the code must import also the components (PR#377)
- - Bug fix: Fixed PATHs handling in glidein_startup.sh (PR#379)
- - Bug fix: Fixed match policy_file import failure (Issue#378, PR#380)
- - Bug fix: Fixed syntax error in ClassAd used for gangliad configuration (Issue#368, PR#385)
- - Bug fix: Added extra logging to investigate file rotation problem (Issue#362, PR#389)
- - Bug fix: Fixed writing of missing dict files during upgrade (Issue#388, PR#391)
- - NOTE-FACTORY: The submit attributes (submit/submit_attrs) are now published in the glidefactory classad with the GlideinSubmit prefix followed by the attribute name and same value. If the attribute name starts with "+" this will be replaced by "_PLUS_", since only alphanumeric characters and "_" are valid in ClassAd attribute names.
-
-
--
- v3_10_5 released on September 27, 2023 (Manual,
- Installation instructions)
-
- - Bug fix: Workaround for EL7 PyJWT bug, generating bytes instead of str (PR#355)
- - Bug fix: Fixed missing `cvmfsexec.cfg` files from Factory reconfig and improved cvmfsexec warnings (Issue#348, PR#356)
- - Bug fix: Added bash requirement to files using bashisms, notably `glidein_sitewms_setup.sh` (PR#358)
- - Bug fix: Fixed syntax errors in analyze_queues (PR#357)
- - Bug fix: Fixed setup_x509 to be successful w/ TRUST_DOMAIN set in the as Factory or Frontend parameter (PR#359)
- - Bug fix: GLIDEIN_SINGULARITY_BINARY_OVERRIDE set also with Frontend and Factory params, not only WN environment (PR#360)
-
-
--
- v3_10_4 released on September 14, 2023 (Manual,
- Installation instructions)
-
- - Bug fix: Fixed missing arguments from rrdtool fetch call (Issue#351, PR#352)
- - Bug fix: gconfig.py to use `gwms-python`, not use `python3` (Issue#349, PR#350)
- - Bug fix: Fixed alternative Shell code still using the `python` (i.e. python2) interpreter (Issue#289, PR#353)
-
-
--
- v3_10_3 released on September 11, 2023 (Manual,
- Installation instructions)
-
- - Added support for Debian 11 and Ubuntu 22 worker nodes (PR#320)
- - Added structured logging. It is a hybrid format with some fields followed by a JSON dictionary. The exact format of the messages may change in the future, and we plan for it to become the default. Now it is disabled by default. Add `structured="True"` to all `process_log` elements (PR#333)
- - Add option to set xml output directory in OSG_autoconf (PR#319)
- - Allow OSG_autoconf to skip sites or CEs that are not present in the OSG collector (PR#315)
- - Add option to set num_factories in OSG_autoconf (Issue#344, PR#345)
- - Added the ability to clear a list of variables from the environment via GLIDEIN_CONTAINER_ENV_CLEARLIST before starting a container (Issue#341, PR#342)
- - Bug fix: Removed `classad` from requirements.txt. The HTCSS team distributes only the `htcondor` library in PyPI which includes a different version of classad (PR#301)
- - Bug fix: Fixing Python 3.9 deprecations (`imp`, `getchildren()` in `xml.etree.ElementTree`) (PR#302, PR#303)
- - Bug fix: Populate missing Entry parameters for ARC CEs submissions (PR#304)
- - Bug fix: Modified the usage of subprocess module, for building/rebuilding cvmfsexec distributions, only when necessary (PR#309)
- - Bug fix: manual_glidein_submit now correctly sets idtokens in the EncryptedInputFiles (issue#283, PR#284)
- - Bug fix: Fixed fetch_rrd crash in EL9 causing missing monitoring and glidefactoryclient classad information (Issue#338, PR#339)
- - NOTE-FACTORY: When generating cvmfsexec distribution for EL9 machine type on an EL7 machine, the factory reconfig and/or upgrade fails as a result of an error in `create_cvmfsexec_distros.sh`. This is possibly due to the tools for EL7 being unable to handle EL9 files (as per Dave Dykstra). Please exercise caution if using `rhel9-x86_64` in the `mtypes` list for the `cvmfsexec_distro` tag in factory configuration. - Our workaround is to remove the EL9 machine type from the default list of machine types supported by the custom distros creation script. Add it back if you are running on an EL9 system and want an EL9 cvmfsexec distrinution. (PR#312)'
-
-
--
- v3_10_2 released on May 10, 2023 (Manual,
- Installation instructions)
-
- - Add a parameter to control the name of the keyname for idtokens (PR#268)
- - Added a factory knob to allow control over rebuilding of cvmfsexec distributions (PR#272)
- - RHEL9 worker node are now recognized by condor_platform_select automatic OS detection (PR#285)
- - Removed pre-reconfigure hook used for rebuilding cvmfsexec distributions whenever a factory reconfig/upgrade was run (issue#262)
- - Bug fix: Use correct variable for `$exit_code` in `singularity_exec_simple` (PR#259)
- - Bug fix: Make sure default value does not overwrite the global one for idtoken_lifetime (PR#261)
- - Bug fix: Protect OSG_autoconf from OSG collector unavailability (PR#276)
- - Bug fix: Fixed jobs going in unknown state in factory monitoring. added QUEUING state for new ARC-CEs REST (PR#286)
- - NOTE: When using on-demand CVMFS, all Glideins after the first one on a node are failing (issue#287)
- - NOTE-FACTORY: The rebuilding of cvmfsexec distributions is now disabled by default, it must be enabled via the new factory knob.\ If you are upgrading from v3.9.5 or later, check that the cvmfsexec script is no more in the pre-reconfigure hooks directory.
-
-
--
- v3_10_1 released on December 13, 2022 (Manual,
- Installation instructions)
-
- - Added utility function to replace error_gen in python scripts (PR#254)
- - Bug fix: Fixed status reporting for `gconfig.py` (PR#254)
- - NOTE: Please see all the notes in 3.9.6 about the custom script changes
-
-
--
- v3_10_0 released on December 7, 2022 (Manual,
- Installation instructions)
-
- - All the features and fixes in v3_9_6
- - Use `SINGULARITY_DISABLE_PID_NAMESPACES` disable `--pid` in Singularity/Apptainer (OSG SOFTWARE-5340, PR#232)
- - Raise a warning if there are "FATAL" errors in Singularity/Apptainer stdlog and the exit code is 0 (PR#235)
- - Added `gconfig.py`. Python utilities to read and write glidein_config (PR#237)
- - Bug fix: Set PATH to default value instead of emptying it (PR#233)
- - Bug fix: Have `get_prop_str` return the default value when the attribute is "undefined" and `gwms_from_config` return the default when set instead of an empty string (PR#235)
- - Bug fix: Fixed credential ID in Glideins. It was not set for scitokens causing incorrect monitoring values (PR#242)
- - Bug fix: Fixed typo in `singularity_lib.sh` (issue#249)
- - NOTE: This follows v3_9_6. Please see all the notes about the custom script changes
-
-
--
- v3_6_5 released on October 5, 2020 (Manual,
- Installation instructions)
-
- - Added GitHub actions to test the code (Python unit tests, BARS, pylint, pycodestyle)
- - Bug fix: Fix for paths in Singularity not substituted correctly when the execute directory has a link or bind mount
- - Bug fix: Fixed shared_port_port to adapt to HTCondor 8.8.10 changes
- - Bug fix: Fixed overrides in OSG_autoconf using depcopy, use https for gfdiff
- - Bug fix: Fixed credential dir in manual_glidein_submit removing hardcoding
-
-
--
- v3_6_4 released on September 17, 2020 (Manual,
- Installation instructions)
-
- - Improved OSG_autoconf
- - Dropped Python 2.6 support
- - Added support for RHEL 8 worker nodes
- - Harmonized, improved and extended CI tests (same CLI, support for Python 3, ...)
- - Added tool to verify if the code for a ticket has been merged
- - Bug fix: Fix for condor_chirp not finding htchirp in the PYTHONPATH
- - Bug fix: Fixed unquoted SINGULARITY_PATH causing HTCondor not to start
-
-
--
- v3_6_3 released on July 16, 2020 (Manual,
- Installation instructions,Tarball)
-
- - Add Python condor_chirp to PATH
- - Dropped tarball installation support
- - Add Python condor_chirp and other GWMS tools to PATH in Singularity
- - Removed "Group:System Environment/Daemons" from glideinwms.spec in OSG repo
- - Added new API documentation using sphinx
- - Added a runner to run the different tests with consistent CLI
- - Added new RHEL 8 Singularity image
- - Updated the singularity wrapper with new features form the OSG wrapper
- - Clean PYTHONPATH and setup GWMS environment in Singularity containers
- - Improved the environment handling of the Singularity wrapper
- - Bug fix: Upgrade the factory documentation not to ask to install the switchboard
- - Bug fix: The Python version of condor_chirp now works with Singularity jobs
- - Bug fix: Enable RFC proxies by default for direct proxies
- - Bug fix: Fix OSG_autoconf and gfdiff
- - Bug fix: Added COMPLETED to the known list of "GridJobStatus"es
- - Bug fix: Add a knob to set entry limits in OSG_autoconf
- - Bug fix: Stop pilots even after the downtime timestamp has been reached when draining pilots
- - Bug fix: Glideins are unrecoverable unless listed as recoverable
- - Bug fix: User job invocation via singularity eating up -e or -n
-
-
--
- v3_6_2 released on March 26, 2020 (Manual,
- Installation instructions,Tarball)
-
- - Add a portable condor_chirp for user jobs running in the glideins
- - Automate the generation of factory configuration
- - Improve control of setup of module and spack in Singularity using MODULE_USE
- - Adding GPU monitor as default for pilots
- - Reduce number of queries from Frontend to User collector
- - Make lgwms-logcat.sh work with non standard factory instance names
- - Bug fix: Pilot using an expired proxy when authenticating to the collector
- - Bug fix: CI issues (coverage broken in SL6, script selection affected by file name)
- - Bug fix: Incorrect CERTIFICATE_MAPFILE used when using HTCondor python binding (failed schedd authentication)
- - Bug fix: Fix release manager relative imports
- - Bug fix: Broken submission to GCE and AWS
- - Bug fix: Correct how entry_ls detect an rpm installation
- - Bug fix: Adjust size of ClusterId and ProcessId fields for the condor_q custom format
- - Bug fix: Total of frontend info is 0 in Factory monitring
-
-
--
- v3_6_1 released on Novemver 13, 2019 (Manual,
- Installation instructions,Tarball)
-
- - Added compatibility w/ HTCondor 8.8.x in OSG 3.5
- - Monitoring pages use https if available
- - Improved search and testing of Singularity binary
- - Unset LD_LIBRARY_PATH and PATH for jobs in Singularity
- - Updated documentation links and Google search
- - Improved CI testing
- - Stop considering held limits when counting maximum jobs in Factory
- - Bug fix: Fix Fatcory tools (entry_rm, entry_q and entry_ls) to be more verbose with single user Factory
- - Bug fix: Removed hardcoded CVMFS requirement for Singularity
- - Bug fix: Improve diagnostic messages when rsa.key file is corrupted
- - Bug fix: Improve control of the worker node draining, avoiding to stop accepting jobs as soon as the draining file is present
-
-
--
- v3_6 released on September 25, 2019 (Manual,
- Installation instructions,Tarball)
-
- - Renaming of v3_5_1 (released on 9/18) to respect the odd/even numbering
- - Includes all 3.4.6 features
- - Updated SW and docs for the change in OSG factories
- - Updated all the obsolete links to HTCondor manual in GlideinWMS website
- - Set up an ITB Frontend for GWMS and FIFE testing
- - Updated gitattributes to resolve conflicts w/ checksum files
- - Added editorconfig and default encoding
- - GlideinWMS code will use now Google docstring format
- - Advertise if a Glidein can use privileged or unprivileged Singularity
- - Check if single user factory migration script has been run before startup
- - Bug fix: pip errors in nightly CI
- - Bug fix: Unittest failing at times on SL7
- - Bug fix: Factory could start also w/ a GT2 entry enabled
- - NOTE: v3_6 is fully interoperable w/ v3_5 and v3_4. I.e. You can mix v3_6_x, v3_5_x and v3_4_x Factories and Frontends in your system
- - NOTE-FACTORY: The v3.6 Factory (like v3.5) is using a single user. This is a major change. If you are upgrading, after the RPM upgrade you will need to:\ 1. stop Factory and HTCondor, 2. Migrate the HTCondor status running the fact_chown script, 3. Run upgrade and restart the HTCondor and the Factory. For detailed instructions see http://glideinwms.fnal.gov/doc.dev/factory/configuration.html#single_user
-
-
-
-
- v3_4_6 released on August 14, 2019 (v3_10_8 released on November 21, 2024 (Manual,,
+ Installation instructions,Tarball)
-
- Make Factory compatible with older 3.4 Frontends reverting back
- to send voms and glexec attributes as strings
+ Advertising information about unprivileged user namespaces in
+ glidein classad (PR#416)
+ - Added option --group-name option to manual_glidein_submit
-
- Backport: GlideinWMS proxy renewal service broken for Xenon
+ Bug fix: Fixed root unable to remove other users jobs in the
+ Factory (PR#433)
-
- Add always --contain to the Singularity invocation and update
- wrapper adding improvement in the OSG one
+ Bug fix: HTCondor TRUST_DOMAIN configuration macro set to string
+ to avoid Glidein config error (PR#420)
- - Document and expand multi-node Glidein
- - Document usage of ignore_down_entries
-
- Bug fix: Not escaped comma in GSI_DAEMON_NAME causing problems
+ Bug fix: Disabled shebang mangling in rpm_build to avoid
+ gwms-python not finding the shell (Issue#436, PR#437)
-
- Bug fix: Debug options causing Singularity invocation to fail
+ Bug fix: Dynamic creation of HTCondor IDTOKEN password
+ (Issue#440, PR#441)
-
- Bug fix: Singularity workdir creation failing at some sites
+ Bug fix: Autodetect CONDOR_OS in the manual_glidein_submit tool
+ (Issue#449, PR#453)
-
- Bug fix: Fixing chkconfig lines on proxy renewal (SOFTWARE-3758)
+ Bug fix: Failed log rotation due to wrong file creation time
+ (Issue#451, PR#457)
-
-
- -
- v3_4_5 released on April 19, 2019 (Manual,Installation instructions,Tarball)
-
- - Add a bind mount for /etc/localtime in Singularity
-
- Bug fix: Strings evaluated as boolean (New error preventing the
- Frontend fo match jobs)
+ Bug fix: Replacing xmlrunner with unittest-xml-reporting
+ (PR#428)
+
+ -
+ Bug fix: Updated the release upload script to work with
+ osg-sw-submit (PR#439)
-
- v3_4_4 released on April 4, 2019 (v3_10_7 released on June 21, 2024 (Manual,,
+ Installation instructions,Tarball)
-
- Propagate to Factory and glidein submission attributes
- controlled by FE
-
- - Track jobs that spawn multiple nodes, e.g. HPC submission
- -
- Include OSG distributed unprivileged Singularity to the search
- path and do a full test of Singularity
+ Apptainer cache and temporary directory set in the Glidein
+ working directory (Issue#403, PR#404)
-
- Improved HTCondor configuration for schedd in Factory and
- Frontend
+ Ability to set a minimum required memory for partitionable
+ Glideins. The default is the value used previously, 2500 MB
+ (Issue#405, PR#406)
- - Add a bind mount for /etc/hosts in Singularity
- - Removed Frontend service dependency from HTCondor service
-
- Added flag to Ignore entries in downtime when considering
- glideins matches
+ Blackhole Detection. Stop accepting jobs if they are consumed at
+ a rate higher than the configured limit and declare the Glidein
+ a blackhole (Issue#331, PR#399)
-
- Bug fix: Factory GlideinMonitor* classads appear to be erased
- periodically
+ Bug fix: Fixed Apptainer validation not considering uid_map w/o
+ initial blank (Issue#395, PR#396)
- - Bug fix: Glidein not killing condor processes
- - Bug fix: Error preventing the Frontend fo match jobs
- - Bug fix: Make sure metasite limits are respected
-
- Bug fix: Convert per/frontend limits to integers before division
+ Bug fix: Flattening error message in \_CONDOR_WRAPPER_ERROR_FILE
+ and JobWrapperFailure Ad. This is a workaround for a bug in
+ HTCondor (PR#400)
- - Bug fix: 'tochild' AttributeError in Factory
-
- Bug fix: Pilot proxies could be generated with invalid VOMS ACs
+ Bug fix: Fixed problem when check_signature in glidein_startup
+ is not defined (PR#402)
-
- NOTE: To use the new HTCondor configuration you need to reconfig
- or restart HTCondor after merging eventual .rpmnew files
+ Bug fix: get_tarballs look for HTCondor releases also in the
+ update directory (Issue#412, PR#413)
-
- v3_4_3 released on January 25, 2019 (v3_10_6 released on January 25, 2024 (Manual,,
+ Installation instructions,Tarball)
-
- Add a scaling factor for all glideins limits in the entries
-
- -
- Comprehensive unit-test for the parsing of the Frontend
- configuration with collectors using shared port
-
- - Add the possibility to disable completely Glidein removal
- - Increase verbosity to help Singularity troubleshooting
- - Improve error messages in glideFactoryCredentials
- - Automatically remove glideins after walltime
- - Bug fix: Improve manual_submit_glidein
- -
- Bug fix: Improve the way condor_jdl dict is populated for
- metasites
+ Add knobs to allow overloading of memory,
+ GLIDEIN_OVERLOAD_MEMORY, and CPU, GLIDEIN_OVERLOAD_CPUS.
+ (Issue#370, PR#374)
+ - Added HTCondor tarball downloader (Issue#367, PR#366)
-
- Bug fix: Factory crashing with malformed HTCondor log
- (AttributeError: dirSummaryTimingsOut.data)
+ Added default (/bin,/usr/bin) when PATH is empty in
+ glidein_startup.sh (PR#373)
-
- Bug fix: glidein_config parsing cutting the results at the first
- space
+ Advertising Factory's HTCondor submit parameters (Issue#307,
+ PR#382)
-
- Bug fix: shell scripts improvements (condition
- smart_partionable.sh, variable splitting)
+ Bug fix: Changed M2Crypto imports to be compatible with 0.40.0
+ the code must import also the components (PR#377)
-
- Bug fix: Frontend upgrade is failing if it is unable to
- determine the version of the Factory
+ Bug fix: Fixed PATHs handling in glidein_startup.sh (PR#379)
- - Bug fix: Avoid glideFactoryEntryGroup processe leaks
-
- Bug fix: Stale running and held glidein numbers reported in
- factory classads
+ Bug fix: Fixed match policy_file import failure (Issue#378,
+ PR#380)
-
- Bug fix: Update documentation about ports requirement for
- Frontend's submit host
+ Bug fix: Fixed syntax error in ClassAd used for gangliad
+ configuration (Issue#368, PR#385)
-
- Bug fix: The factory seems to ignore the configuration values in
- the files in the config.d directory w/ entry configurations
+ Bug fix: Added extra logging to investigate file rotation
+ problem (Issue#362, PR#389)
- - Bug fix: Frontend not recognizing entries in downtime
-
- Bug fix: Pylint tests improved (htcondor setup, remove external
- lins)
+ Bug fix: Fixed writing of missing dict files during upgrade
+ (Issue#388, PR#391)
-
- NOTE: If you update from 3.4 or earlier make sure to have open
- port 9618 also on standalone Frontends/subit hosts (was using
- 9615 before)
+ NOTE-FACTORY: The submit attributes (submit/submit_attrs) are
+ now published in the glidefactory classad with the GlideinSubmit
+ prefix followed by the attribute name and same value. If the
+ attribute name starts with "+" this will be replaced by
+ "_PLUS_", since only alphanumeric characters and "_" are valid
+ in ClassAd attribute names.
-
- v3_4_2 released on October 26, 2018 (v3_10_5 released on September 27, 2023
+ (Manual,,
+ Installation instructions,Tarball)
-
- Controlling that Frontend is not using options incompatible w/
- linked Factories
+ Bug fix: Workaround for EL7 PyJWT bug, generating bytes instead
+ of str (PR#355)
+
+ -
+ Bug fix: Fixed missing `cvmfsexec.cfg` files from Factory
+ reconfig and improved cvmfsexec warnings (Issue#348, PR#356)
+
+ -
+ Bug fix: Added bash requirement to files using bashisms, notably
+ `glidein_sitewms_setup.sh` (PR#358)
+
+ - Bug fix: Fixed syntax errors in analyze_queues (PR#357)
+ -
+ Bug fix: Fixed setup_x509 to be successful w/ TRUST_DOMAIN set
+ in the as Factory or Frontend parameter (PR#359)
+
+ -
+ Bug fix: GLIDEIN_SINGULARITY_BINARY_OVERRIDE set also with
+ Frontend and Factory params, not only WN environment (PR#360)
- - Use systemctl for loading/unloading on EL7
-
- v3_4_1 released on October 18, 2018 (v3_10_4 released on September 14, 2023
+ (Manual,,
+ Installation instructions,Tarball)
-
- Improved Singularity support (improved negotiation, added bind
- path variables and GPU support)
-
- - Switch User collector child collectors to shared_port
- -
- Updated documentation: remove obsolete parts, verified all
- references
+ Bug fix: Fixed missing arguments from rrdtool fetch call
+ (Issue#351, PR#352)
- - Increased unit test coverage
- - Added the possibility to skip idle removal per entry
- - Improved how subentries are picked for metasites
-
- Bug fix: Periodic scripts fails but report GLIDEIN_PS_OK=True
+ Bug fix: gconfig.py to use `gwms-python`, not use `python3`
+ (Issue#349, PR#350)
-
- Bug fix: Fix entry_set configuration not to load temporary files
- and to work across upgrades
+ Bug fix: Fixed alternative Shell code still using the `python`
+ (i.e. python2) interpreter (Issue#289, PR#353)
- - Big fix: fixed metasites monitoring
-
- v3_4 released on June 4, 2018 (v3_10_3 released on September 11, 2023
+ (Manual,,
+ Installation instructions,Tarball)
-
- GlideinWMS 3.4 includes all features of previous stable series
- 3.2 and development series 3.3
+ Added support for Debian 11 and Ubuntu 22 worker nodes (PR#320)
-
- Inform the admin in case multiple service reload is done in
- rapid succession
+ Added structured logging. It is a hybrid format with some fields
+ followed by a JSON dictionary. The exact format of the messages
+ may change in the future, and we plan for it to become the
+ default. Now it is disabled by default. Add `structured="True"`
+ to all `process_log` elements (PR#333)
-
- Released glideinwms-switchdoard RPM to provide
- condor_root_switchboard
+ Add option to set xml output directory in OSG_autoconf (PR#319)
-
- Singularity wrapper using PATH and module when SINGULARITY_BIN
- does not contain the correct path
+ Allow OSG_autoconf to skip sites or CEs that are not present in
+ the OSG collector (PR#315)
- - Update of the software license
-
- Have frontend groups glidein requests track the job requests
+ Add option to set num_factories in OSG_autoconf (Issue#344,
+ PR#345)
-
- Add a resource_slot that adds CPUs to the system but is in the
- main slot
+ Added the ability to clear a list of variables from the
+ environment via GLIDEIN_CONTAINER_ENV_CLEARLIST before starting
+ a container (Issue#341, PR#342)
-
- singularity jobs each use a separate linux session to support
- restricted-access CVMFS
+ Bug fix: Removed `classad` from requirements.txt. The HTCSS team
+ distributes only the `htcondor` library in PyPI which includes a
+ different version of classad (PR#301)
-
- Do not set GLIDEIN_ToDie based on X509 user proxy expiration
+ Bug fix: Fixing Python 3.9 deprecations (`imp`, `getchildren()`
+ in `xml.etree.ElementTree`) (PR#302, PR#303)
-
- Estimate the cores provided to glideins running on an entry
+ Bug fix: Populate missing Entry parameters for ARC CEs
+ submissions (PR#304)
- - Write LogCompletedStats jsons for monitoring purposes
- - Add entry monitoring breakdown for metasites
-
- Code modernization to Python 2.6/2.7 stansards (futurize stage 1
- compliant)
+ Bug fix: Modified the usage of subprocess module, for
+ building/rebuilding cvmfsexec distributions, only when necessary
+ (PR#309)
-
- Increased testing (more than doubled unit tests coverage and
- testing more files w/ pylint)
+ Bug fix: manual_glidein_submit now correctly sets idtokens in
+ the EncryptedInputFiles (issue#283, PR#284)
-
- Bug Fix: TypeError: under special conditions in Factory reconfig
+ Bug fix: Fixed fetch_rrd crash in EL9 causing missing monitoring
+ and glidefactoryclient classad information (Issue#338, PR#339)
-
- Bug Fix: Entries in downtime slow down the ramp-up for the
- frontend
+ NOTE-FACTORY: When generating cvmfsexec distribution for EL9
+ machine type on an EL7 machine, the factory reconfig and/or
+ upgrade fails as a result of an error in
+ `create_cvmfsexec_distros.sh`. This is possibly due to the tools
+ for EL7 being unable to handle EL9 files (as per Dave Dykstra).
+ Please exercise caution if using `rhel9-x86_64` in the `mtypes`
+ list for the `cvmfsexec_distro` tag in factory configuration. -
+ Our workaround is to remove the EL9 machine type from the
+ default list of machine types supported by the custom distros
+ creation script. Add it back if you are running on an EL9 system
+ and want an EL9 cvmfsexec distrinution. (PR#312)'
+
+
+
+ -
+ v3_10_2 released on May 10, 2023 (Manual,
+ Installation instructions)
+
+ -
+ Add a parameter to control the name of the keyname for idtokens
+ (PR#268)
-
- NOTE: The type of the GLIDEIN_CPU attr is String (to accomodate
- the keywords auto, slot, node). Documentation was reporting Int
- incorrectly. Make sure your configuration uses the correct type
- or you may get a reconfig/upgrade error.
+ Added a factory knob to allow control over rebuilding of
+ cvmfsexec distributions (PR#272)
-
- NOTE: If you use HTCondor 8.7.2 or bigger with the GlideinWMS
- Factory, you must install also glideinwms-switchboard, to have
- condor_root_switchboard, that was removed from HTCondor
+ RHEL9 worker node are now recognized by condor_platform_select
+ automatic OS detection (PR#285)
-
- NOTE: 'entry_sets' should be considered an experimental feature:
- the imlpementation is changing and there may be errors when
- upgrading across versions
+ Removed pre-reconfigure hook used for rebuilding cvmfsexec
+ distributions whenever a factory reconfig/upgrade was run
+ (issue#262)
+
+ -
+ Bug fix: Use correct variable for `$exit_code` in
+ `singularity_exec_simple` (PR#259)
+
+ -
+ Bug fix: Make sure default value does not overwrite the global
+ one for idtoken_lifetime (PR#261)
+
+ -
+ Bug fix: Protect OSG_autoconf from OSG collector unavailability
+ (PR#276)
+
+ -
+ Bug fix: Fixed jobs going in unknown state in factory
+ monitoring. added QUEUING state for new ARC-CEs REST (PR#286)
+
+ -
+ NOTE: When using on-demand CVMFS, all Glideins after the first
+ one on a node are failing (issue#287)
+
+ -
+ NOTE-FACTORY: The rebuilding of cvmfsexec distributions is now
+ disabled by default, it must be enabled via the new factory
+ knob.\ If you are upgrading from v3.9.5 or later, check that the
+ cvmfsexec script is no more in the pre-reconfigure hooks
+ directory.
-
- v3_2_22_2 released on April 17, 2018 (v3_10_1 released on December 13, 2022 (Manual,,
+ Installation instructions,Tarball)
-
- Changed singularity options also in singularity_setup.sh to
- adapt to the new singularity release requirements
+ Added utility function to replace error_gen in python scripts
+ (PR#254)
+ - Bug fix: Fixed status reporting for `gconfig.py` (PR#254)
-
- Bug Fix: Fix bug due to malformed -order option (SOFTWARE-3163)
+ NOTE: Please see all the notes in 3.9.6 about the custom script
+ changes
-
- v3_2_22_1 released on April 11, 2018 (v3_10_0 released on December 7, 2022 (Manual,,
+ Installation instructions,Tarball)
+ - All the features and fixes in v3_9_6
-
- Changed singularity options to adapt to the new singularity
- release requirements
+ Use `SINGULARITY_DISABLE_PID_NAMESPACES` disable `--pid` in
+ Singularity/Apptainer (OSG SOFTWARE-5340, PR#232)
+
+ -
+ Raise a warning if there are "FATAL" errors in
+ Singularity/Apptainer stdlog and the exit code is 0 (PR#235)
+
+ -
+ Added `gconfig.py`. Python utilities to read and write
+ glidein_config (PR#237)
+
+ -
+ Bug fix: Set PATH to default value instead of emptying it
+ (PR#233)
+
+ -
+ Bug fix: Have `get_prop_str` return the default value when the
+ attribute is "undefined" and `gwms_from_config` return the
+ default when set instead of an empty string (PR#235)
+
+ -
+ Bug fix: Fixed credential ID in Glideins. It was not set for
+ scitokens causing incorrect monitoring values (PR#242)
+
+ - Bug fix: Fixed typo in `singularity_lib.sh` (issue#249)
+ -
+ NOTE: This follows v3_9_6. Please see all the notes about the
+ custom script changes
-
- v3_2_22 released on April 10, 2018 (v3_6_5 released on October 5, 2020 (Manual,,
+ Installation instructions,Tarball)
- - Bug Fix: Incorrect behavior of Singularity
- - Bug Fix: proxy-renewal-script updates and bug fixes
-
- Bug Fix: Critical bug in 3.2.21 leads to leaking
- glideFactoryEntryGroup.py processes
+ Added GitHub actions to test the code (Python unit tests, BARS,
+ pylint, pycodestyle)
+
+ -
+ Bug fix: Fix for paths in Singularity not substituted correctly
+ when the execute directory has a link or bind mount
+
+ -
+ Bug fix: Fixed shared_port_port to adapt to HTCondor 8.8.10
+ changes
+
+ -
+ Bug fix: Fixed overrides in OSG_autoconf using depcopy, use
+ https for gfdiff
+
+ -
+ Bug fix: Fixed credential dir in manual_glidein_submit removing
+ hardcoding
-
- v3_2_21 released on February 7, 2018 (v3_6_4 released on September 17, 2020 (Manual,,
+ Installation instructions,Tarball)
+ - Improved OSG_autoconf
+ - Dropped Python 2.6 support
+ - Added support for RHEL 8 worker nodes
-
- Have frontend groups request the removal of unused glideins
+ Harmonized, improved and extended CI tests (same CLI, support
+ for Python 3, ...)
-
- Support of unprivileged singularity and new singularity scripts
+ Added tool to verify if the code for a ticket has been merged
- - Automatically renew gwms proxies
- - Factory monitoring displaying correctly core counters
-
- Balancing glidein pressure to sites that are aliases or
- Meta-Sites
+ Bug fix: Fix for condor_chirp not finding htchirp in the
+ PYTHONPATH
+
+ -
+ Bug fix: Fixed unquoted SINGULARITY_PATH causing HTCondor not to
+ start
- - Remove osg-version requirements
- - Bug Fix: Fix fork.py behavior
- - Bug Fix: Uninitialized variable caused skipping fix-rrd
- - Bug Fix: Sanitize content of MJF attributes
- - Bug Fix: Update Google custom search to new API
-
- v3_2_20 released on November 15, 2017 (v3_6_3 released on July 16, 2020 (Manual,,
+ Installation instructions,,Tarball)
+ - Add Python condor_chirp to PATH
+ - Dropped tarball installation support
-
- NOTE: during the first "upgrade" or "reconfig --fix-rrd" you
- will see some warning because of missing RRD attributes. This is
- OK, is caused by the schema change.
+ Add Python condor_chirp and other GWMS tools to PATH in
+ Singularity
- - Verified that Factory can start 1500 entries at a time
-
- Improved Factory monitoring by adding cores count for running
- and idle jobs and requests
+ Removed "Group:System Environment/Daemons" from glideinwms.spec
+ in OSG repo
- - Removed dependency from Globus clients
- - Support Singularity (future replacement for glexec)
+ - Added new API documentation using sphinx
-
- Avoid race conditions with multiple reload quick invocations in
- SL7
+ Added a runner to run the different tests with consistent CLI
+ - Added new RHEL 8 Singularity image
-
- Bug Fix: -fix_rrd is always in the upgrade command beside
- reconfig option
+ Updated the singularity wrapper with new features form the OSG
+ wrapper
- - Bug Fix: Fixed some failing unit tests
- - Bug Fix: Fixed Factory job stats are empty
- - Bug Fix: Bad link to Frontend monitoring
-
- Bug Fix: Custom Google search in the documentation pointing to
- the new site, not the old mirror on uscms.org
+ Clean PYTHONPATH and setup GWMS environment in Singularity
+ containers
- - Bug Fix: Fix DC_DAEMON_LIST
-
- Bug Fix: verifyRRD hard coded paths, missing some files during
- -fix_rr
+ Improved the environment handling of the Singularity wrapper
-
- Bug Fix: HTCondor QEdit triggered also when
- advertise_pilot_accounting is not set
+ Bug fix: Upgrade the factory documentation not to ask to install
+ the switchboard
-
- Bug Fix: Downtime setting correctly distinguish daylight saving
+ Bug fix: The Python version of condor_chirp now works with
+ Singularity jobs
+
+ - Bug fix: Enable RFC proxies by default for direct proxies
+ - Bug fix: Fix OSG_autoconf and gfdiff
+ -
+ Bug fix: Added COMPLETED to the known list of "GridJobStatus"es
+
+ - Bug fix: Add a knob to set entry limits in OSG_autoconf
+ -
+ Bug fix: Stop pilots even after the downtime timestamp has been
+ reached when draining pilots
+
+ -
+ Bug fix: Glideins are unrecoverable unless listed as recoverable
+
+ -
+ Bug fix: User job invocation via singularity eating up -e or -n
- - Bug Fix: Added usercollector RPM dependency from ganglia
-
- v3_2_19 released on May 30, 2017 (v3_6_2 released on March 26, 2020 (Manual,,
+ Installation instructions,,Tarball)
-
- NOTE: If using HTCondor 8.6.x you have to disable the collector
- use of shared port. In
- /etc/condor/config.d/01_gwms_factory_collectors.config (factory)
- /etc/condor/config.d/01_gwms_collectors.config (frontend, wms
- user collector) and (after the line
- COLLECTOR_USES_SHARED_PORT=False) add the line:
- COLLECTOR.USE_SHARED_PORT=False
+ Add a portable condor_chirp for user jobs running in the
+ glideins
-
- - Added counters for Idle jobs older than X hours
+ - Automate the generation of factory configuration
-
- Enabling GWMS_XSLT_PLUGIN_DIR by default for
- glideinwms-vofrontend-standalone rpm
+ Improve control of setup of module and spack in Singularity
+ using MODULE_USE
- - Linked Frontend monitoring from Factory monitoring
+ - Adding GPU monitor as default for pilots
+ - Reduce number of queries from Frontend to User collector
-
- Improved glideins scale down by adding a timeout,
- GLIDEIN_IDLE_LIFETIME
+ Make lgwms-logcat.sh work with non standard factory instance
+ names
- - Log number of activation/claims per glidein
-
- Several documentation improvements, including reconfig and
- upgrade operations in SL7 RPM installation
+ Bug fix: Pilot using an expired proxy when authenticating to the
+ collector
-
- Bug Fix: reduced the number of file descriptors used per Entry
- in the Factory
+ Bug fix: CI issues (coverage broken in SL6, script selection
+ affected by file name)
-
- Bug Fix: Factory entries submitting glideins even after hitting
- the limit
+ Bug fix: Incorrect CERTIFICATE_MAPFILE used when using HTCondor
+ python binding (failed schedd authentication)
+ - Bug fix: Fix release manager relative imports
+ - Bug fix: Broken submission to GCE and AWS
+ - Bug fix: Correct how entry_ls detect an rpm installation
-
- Bug Fix: Errors with HTCondor 8.5/6 upgrade: SUBSYS.LOCALNAME.*
- warning triggered by GWMS htcondor configuration, Glidein sent
- unintentionally to multiple schedds on the Factory
+ Bug fix: Adjust size of ClusterId and ProcessId fields for the
+ condor_q custom format
+ - Bug fix: Total of frontend info is 0 in Factory monitring
+
+
+ -
+ v3_6_1 released on Novemver 13, 2019 (Manual,
+ Installation instructions,Tarball)
+
+ - Added compatibility w/ HTCondor 8.8.x in OSG 3.5
+ - Monitoring pages use https if available
+ - Improved search and testing of Singularity binary
+ - Unset LD_LIBRARY_PATH and PATH for jobs in Singularity
+ - Updated documentation links and Google search
+ - Improved CI testing
-
- Bug Fix: Clarified attribute types and fixed globbing behavior
+ Stop considering held limits when counting maximum jobs in
+ Factory
-
- Bug Fix: RPM verification fails when config files are changed
+ Bug fix: Fix Fatcory tools (entry_rm, entry_q and entry_ls) to
+ be more verbose with single user Factory
-
- Bug Fix: GLIDEIN_CPU settings "node" and "slot" supported also
- in Frontend, better documented and improved to compensate for
- PBS misconfiguration
+ Bug fix: Removed hardcoded CVMFS requirement for Singularity
-
- Bug Fix: Fixed submission to AWS which uses key_pair as
- auth_method
+ Bug fix: Improve diagnostic messages when rsa.key file is
+ corrupted
+
+ -
+ Bug fix: Improve control of the worker node draining, avoiding
+ to stop accepting jobs as soon as the draining file is present
-
- v3_2_18 released on February 28, 2017 (v3_6 released on September 25, 2019 (Manual,,
+ Installation instructions,,Tarball)
- - Bug Fix: ProjectId is missing double quotes
-
- Bug Fix: Stdout messages from startup script in SL7 are
- confusing
+ Renaming of v3_5_1 (released on 9/18) to respect the odd/even
+ numbering
+ - Includes all 3.4.6 features
+ - Updated SW and docs for the change in OSG factories
-
- Bug Fix: Service definition file for SL7 is marked executable
- and it should not
+ Updated all the obsolete links to HTCondor manual in GlideinWMS
+ website
+ - Set up an ITB Frontend for GWMS and FIFE testing
-
- Bug Fix: Upgrade not working when frontend/factory are running
+ Updated gitattributes to resolve conflicts w/ checksum files
+ - Added editorconfig and default encoding
+ - GlideinWMS code will use now Google docstring format
-
- Bug Fix: Use of daemon function to start process does not play
- well with non empty pid files
+ Advertise if a Glidein can use privileged or unprivileged
+ Singularity
+
+ -
+ Check if single user factory migration script has been run
+ before startup
+
+ - Bug fix: pip errors in nightly CI
+ - Bug fix: Unittest failing at times on SL7
+ - Bug fix: Factory could start also w/ a GT2 entry enabled
+ -
+ NOTE: v3_6 is fully interoperable w/ v3_5 and v3_4. I.e. You can
+ mix v3_6_x, v3_5_x and v3_4_x Factories and Frontends in your
+ system
+
+ -
+ NOTE-FACTORY: The v3.6 Factory (like v3.5) is using a single
+ user. This is a major change. If you are upgrading, after the
+ RPM upgrade you will need to:\ 1. stop Factory and HTCondor, 2.
+ Migrate the HTCondor status running the fact_chown script, 3.
+ Run upgrade and restart the HTCondor and the Factory. For
+ detailed instructions see
+ http://glideinwms.fnal.gov/doc.dev/factory/configuration.html#single_user
+
-
- v3_2_17 released on January 25, 2017 (v3_4_6 released on August 14, 2019 (Manual,,Installation instructions,,Tarball)
-
- Glideins fallback to curl if transferring files from the
- factory/frontend staging area fails using wget
-
-
- -
- Bug Fix: Factory not correctly consider the cluster size when
- doing multiple submissions during same cycle
-
-
- -
- Added support for systemctl init scripts for RHEL6 and RHEL7
+ Make Factory compatible with older 3.4 Frontends reverting back
+ to send voms and glexec attributes as strings
-
- Factory and Frontend service init scripts now use daemon
- functions
+ Backport: GlideinWMS proxy renewal service broken for Xenon
-
- Added ability to request disk space for special resource slots
+ Add always --contain to the Singularity invocation and update
+ wrapper adding improvement in the OSG one
- - Updated jQuery in monitoring webpages to version 1.12
- - Updated Documentation
+ - Document and expand multi-node Glidein
+ - Document usage of ignore_down_entries
-
- Pilot accounting information from jobs is now available in the
- glidein job's classad on the factory side
+ Bug fix: Not escaped comma in GSI_DAEMON_NAME causing problems
-
-
- Frontend service performance stats are now advertised in the
- glidefrontendmonitor classad
+ Bug fix: Debug options causing Singularity invocation to fail
-
-
- Bug Fix: Fixed race condition where periodic scripts in glidein
- would corrupt a glidein_config
+ Bug fix: Singularity workdir creation failing at some sites
-
-
- Bug Fix: Fixed an issue where some START_EXPR were incorrectly
- ignored
+ Bug fix: Fixing chkconfig lines on proxy renewal (SOFTWARE-3758)
-
+
+
+ -
+ v3_4_5 released on April 19, 2019 (Manual,Installation instructions,Tarball)
+
+ - Add a bind mount for /etc/localtime in Singularity
-
- Bug Fix: When not configured GLIDEIN_CPUS default back to 1
- rather than trying to auto detect available cores
+ Bug fix: Strings evaluated as boolean (New error preventing the
+ Frontend fo match jobs)
-
- - Bug Fix: Glideins will not use UTMP be default
-
- v3_2_16 released on October 21, 2016 (v3_4_4 released on April 4, 2019 (Manual,,Installation instructions,,Tarball)
-
- Information about job and machine features are now available in
- glidein startd's classad
+ Propagate to Factory and glidein submission attributes
+ controlled by FE
-
+ - Track jobs that spawn multiple nodes, e.g. HPC submission
-
- New tool manual_glidein_startup now lets you manually start a
- glidein for a given factory entry and frontend group
+ Include OSG distributed unprivileged Singularity to the search
+ path and do a full test of Singularity
-
-
- BOSCO entries can now accept credentials provided from the
- frontend
+ Improved HTCondor configuration for schedd in Factory and
+ Frontend
-
+ - Add a bind mount for /etc/hosts in Singularity
+ - Removed Frontend service dependency from HTCondor service
-
- Bug Fix: Glidein is now correctly auto detect RHEL6 and RHEL7
+ Added flag to Ignore entries in downtime when considering
+ glideins matches
-
- Bug Fix: Dagman and Schedd universe jobs are not counted against
- max jobs running
+ Bug fix: Factory GlideinMonitor* classads appear to be erased
+ periodically
-
+ - Bug fix: Glidein not killing condor processes
+ - Bug fix: Error preventing the Frontend fo match jobs
+ - Bug fix: Make sure metasite limits are respected
-
- Bug Fix: An entry in downtime will now show up in the glidein
- status
+ Bug fix: Convert per/frontend limits to integers before division
-
- - Bug Fix: Added condor-python rpm as a dependency
+ - Bug fix: 'tochild' AttributeError in Factory
-
- Bug Fix: Complex credentials (vm_id+vm_type) are now correctly
- interpreted
+ Bug fix: Pilot proxies could be generated with invalid VOMS ACs
-
-
- Bug Fix: Frontend is now more resilent to transient errors
- communicating with the HTCondor daemons
+ NOTE: To use the new HTCondor configuration you need to reconfig
+ or restart HTCondor after merging eventual .rpmnew files
-
- -
- Bug Fix: Slave frontend now correctly looks up the master
- frontend when using condor-python bindings
+
+
+ -
+ v3_4_3 released on January 25, 2019 (Manual,Installation instructions,Tarball)
+
+ -
+ Add a scaling factor for all glideins limits in the entries
+
+ -
+ Comprehensive unit-test for the parsing of the Frontend
+ configuration with collectors using shared port
+
+ - Add the possibility to disable completely Glidein removal
+ - Increase verbosity to help Singularity troubleshooting
+ - Improve error messages in glideFactoryCredentials
+ - Automatically remove glideins after walltime
+ - Bug fix: Improve manual_submit_glidein
+ -
+ Bug fix: Improve the way condor_jdl dict is populated for
+ metasites
+
+ -
+ Bug fix: Factory crashing with malformed HTCondor log
+ (AttributeError: dirSummaryTimingsOut.data)
+
+ -
+ Bug fix: glidein_config parsing cutting the results at the first
+ space
+
+ -
+ Bug fix: shell scripts improvements (condition
+ smart_partionable.sh, variable splitting)
+
+ -
+ Bug fix: Frontend upgrade is failing if it is unable to
+ determine the version of the Factory
+
+ - Bug fix: Avoid glideFactoryEntryGroup processe leaks
+ -
+ Bug fix: Stale running and held glidein numbers reported in
+ factory classads
+
+ -
+ Bug fix: Update documentation about ports requirement for
+ Frontend's submit host
+
+ -
+ Bug fix: The factory seems to ignore the configuration values in
+ the files in the config.d directory w/ entry configurations
+
+ - Bug fix: Frontend not recognizing entries in downtime
+ -
+ Bug fix: Pylint tests improved (htcondor setup, remove external
+ lins)
+
+ -
+ NOTE: If you update from 3.4 or earlier make sure to have open
+ port 9618 also on standalone Frontends/subit hosts (was using
+ 9615 before)
+
+
+
+ -
+ v3_4_2 released on October 26, 2018 (Manual,Installation instructions,Tarball)
+
+ -
+ Controlling that Frontend is not using options incompatible w/
+ linked Factories
+
+ - Use systemctl for loading/unloading on EL7
+
+
+ -
+ v3_4_1 released on October 18, 2018 (Manual,Installation instructions,Tarball)
+
+ -
+ Improved Singularity support (improved negotiation, added bind
+ path variables and GPU support)
+
+ - Switch User collector child collectors to shared_port
+ -
+ Updated documentation: remove obsolete parts, verified all
+ references
+
+ - Increased unit test coverage
+ - Added the possibility to skip idle removal per entry
+ - Improved how subentries are picked for metasites
+ -
+ Bug fix: Periodic scripts fails but report GLIDEIN_PS_OK=True
+
+ -
+ Bug fix: Fix entry_set configuration not to load temporary files
+ and to work across upgrades
+
+ - Big fix: fixed metasites monitoring
+
+
+ -
+ v3_4 released on June 4, 2018 (Manual,Installation instructions,Tarball)
+
+ -
+ GlideinWMS 3.4 includes all features of previous stable series
+ 3.2 and development series 3.3
+
+ -
+ Inform the admin in case multiple service reload is done in
+ rapid succession
+
+ -
+ Released glideinwms-switchdoard RPM to provide
+ condor_root_switchboard
+
+ -
+ Singularity wrapper using PATH and module when SINGULARITY_BIN
+ does not contain the correct path
+
+ - Update of the software license
+ -
+ Have frontend groups glidein requests track the job requests
+
+ -
+ Add a resource_slot that adds CPUs to the system but is in the
+ main slot
+
+ -
+ singularity jobs each use a separate linux session to support
+ restricted-access CVMFS
+
+ -
+ Do not set GLIDEIN_ToDie based on X509 user proxy expiration
+
+ -
+ Estimate the cores provided to glideins running on an entry
+
+ - Write LogCompletedStats jsons for monitoring purposes
+ - Add entry monitoring breakdown for metasites
+ -
+ Code modernization to Python 2.6/2.7 stansards (futurize stage 1
+ compliant)
+
+ -
+ Increased testing (more than doubled unit tests coverage and
+ testing more files w/ pylint)
+
+ -
+ Bug Fix: TypeError: under special conditions in Factory reconfig
+
+ -
+ Bug Fix: Entries in downtime slow down the ramp-up for the
+ frontend
+
+ -
+ NOTE: The type of the GLIDEIN_CPU attr is String (to accomodate
+ the keywords auto, slot, node). Documentation was reporting Int
+ incorrectly. Make sure your configuration uses the correct type
+ or you may get a reconfig/upgrade error.
+
+ -
+ NOTE: If you use HTCondor 8.7.2 or bigger with the GlideinWMS
+ Factory, you must install also glideinwms-switchboard, to have
+ condor_root_switchboard, that was removed from HTCondor
+
+ -
+ NOTE: 'entry_sets' should be considered an experimental feature:
+ the imlpementation is changing and there may be errors when
+ upgrading across versions
+
+
+
+ -
+ v3_2_22_2 released on April 17, 2018 (Manual,Installation instructions,Tarball)
+
+ -
+ Changed singularity options also in singularity_setup.sh to
+ adapt to the new singularity release requirements
+
+ -
+ Bug Fix: Fix bug due to malformed -order option (SOFTWARE-3163)
+
+
+
+ -
+ v3_2_22_1 released on April 11, 2018 (Manual,Installation instructions,Tarball)
+
+ -
+ Changed singularity options to adapt to the new singularity
+ release requirements
+
+
+
+ -
+ v3_2_22 released on April 10, 2018 (Manual,Installation instructions,Tarball)
+
+ - Bug Fix: Incorrect behavior of Singularity
+ - Bug Fix: proxy-renewal-script updates and bug fixes
+ -
+ Bug Fix: Critical bug in 3.2.21 leads to leaking
+ glideFactoryEntryGroup.py processes
+
+
+
+ -
+ v3_2_21 released on February 7, 2018 (Manual,Installation instructions,Tarball)
+
+ -
+ Have frontend groups request the removal of unused glideins
+
+ -
+ Support of unprivileged singularity and new singularity scripts
+
+ - Automatically renew gwms proxies
+ - Factory monitoring displaying correctly core counters
+ -
+ Balancing glidein pressure to sites that are aliases or
+ Meta-Sites
+
+ - Remove osg-version requirements
+ - Bug Fix: Fix fork.py behavior
+ - Bug Fix: Uninitialized variable caused skipping fix-rrd
+ - Bug Fix: Sanitize content of MJF attributes
+ - Bug Fix: Update Google custom search to new API
+
+
+ -
+ v3_2_20 released on November 15, 2017 (Manual,Installation instructions,Tarball)
+
+ -
+ NOTE: during the first "upgrade" or "reconfig --fix-rrd" you
+ will see some warning because of missing RRD attributes. This is
+ OK, is caused by the schema change.
+
+ - Verified that Factory can start 1500 entries at a time
+ -
+ Improved Factory monitoring by adding cores count for running
+ and idle jobs and requests
+
+ - Removed dependency from Globus clients
+ - Support Singularity (future replacement for glexec)
+ -
+ Avoid race conditions with multiple reload quick invocations in
+ SL7
+
+ -
+ Bug Fix: -fix_rrd is always in the upgrade command beside
+ reconfig option
+
+ - Bug Fix: Fixed some failing unit tests
+ - Bug Fix: Fixed Factory job stats are empty
+ - Bug Fix: Bad link to Frontend monitoring
+ -
+ Bug Fix: Custom Google search in the documentation pointing to
+ the new site, not the old mirror on uscms.org
+
+ - Bug Fix: Fix DC_DAEMON_LIST
+ -
+ Bug Fix: verifyRRD hard coded paths, missing some files during
+ -fix_rr
+
+ -
+ Bug Fix: HTCondor QEdit triggered also when
+ advertise_pilot_accounting is not set
+
+ -
+ Bug Fix: Downtime setting correctly distinguish daylight saving
+
+ - Bug Fix: Added usercollector RPM dependency from ganglia
+
+
+ -
+ v3_2_19 released on May 30, 2017 (Manual,Installation instructions,Tarball)
+
+ -
+ NOTE: If using HTCondor 8.6.x you have to disable the collector
+ use of shared port. In
+ /etc/condor/config.d/01_gwms_factory_collectors.config (factory)
+ /etc/condor/config.d/01_gwms_collectors.config (frontend, wms
+ user collector) and (after the line
+ COLLECTOR_USES_SHARED_PORT=False) add the line:
+ COLLECTOR.USE_SHARED_PORT=False
+
+
+ - Added counters for Idle jobs older than X hours
+ -
+ Enabling GWMS_XSLT_PLUGIN_DIR by default for
+ glideinwms-vofrontend-standalone rpm
+
+ - Linked Frontend monitoring from Factory monitoring
+ -
+ Improved glideins scale down by adding a timeout,
+ GLIDEIN_IDLE_LIFETIME
+
+ - Log number of activation/claims per glidein
+ -
+ Several documentation improvements, including reconfig and
+ upgrade operations in SL7 RPM installation
+
+ -
+ Bug Fix: reduced the number of file descriptors used per Entry
+ in the Factory
+
+ -
+ Bug Fix: Factory entries submitting glideins even after hitting
+ the limit
+
+ -
+ Bug Fix: Errors with HTCondor 8.5/6 upgrade: SUBSYS.LOCALNAME.*
+ warning triggered by GWMS htcondor configuration, Glidein sent
+ unintentionally to multiple schedds on the Factory
+
+ -
+ Bug Fix: Clarified attribute types and fixed globbing behavior
+
+ -
+ Bug Fix: RPM verification fails when config files are changed
+
+ -
+ Bug Fix: GLIDEIN_CPU settings "node" and "slot" supported also
+ in Frontend, better documented and improved to compensate for
+ PBS misconfiguration
+
+ -
+ Bug Fix: Fixed submission to AWS which uses key_pair as
+ auth_method
+
+
+
+ -
+ v3_2_18 released on February 28, 2017 (Manual,Installation instructions,Tarball)
+
+ - Bug Fix: ProjectId is missing double quotes
+ -
+ Bug Fix: Stdout messages from startup script in SL7 are
+ confusing
+
+ -
+ Bug Fix: Service definition file for SL7 is marked executable
+ and it should not
+
+ -
+ Bug Fix: Upgrade not working when frontend/factory are running
+
+ -
+ Bug Fix: Use of daemon function to start process does not play
+ well with non empty pid files
+
+
+
+ -
+ v3_2_17 released on January 25, 2017 (Manual,Installation instructions,Tarball)
+
+ -
+ Glideins fallback to curl if transferring files from the
+ factory/frontend staging area fails using wget
+
+
+ -
+ Bug Fix: Factory not correctly consider the cluster size when
+ doing multiple submissions during same cycle
+
+
+ -
+ Added support for systemctl init scripts for RHEL6 and RHEL7
+
+ -
+ Factory and Frontend service init scripts now use daemon
+ functions
+
+ -
+ Added ability to request disk space for special resource slots
+
+ - Updated jQuery in monitoring webpages to version 1.12
+ - Updated Documentation
+ -
+ Pilot accounting information from jobs is now available in the
+ glidein job's classad on the factory side
+
+
+ -
+ Frontend service performance stats are now advertised in the
+ glidefrontendmonitor classad
+
+
+ -
+ Bug Fix: Fixed race condition where periodic scripts in glidein
+ would corrupt a glidein_config
+
+
+ -
+ Bug Fix: Fixed an issue where some START_EXPR were incorrectly
+ ignored
+
+
+ -
+ Bug Fix: When not configured GLIDEIN_CPUS default back to 1
+ rather than trying to auto detect available cores
+
+
+ - Bug Fix: Glideins will not use UTMP be default
+
+
+ -
+ v3_2_16 released on October 21, 2016 (Manual,Installation instructions,Tarball)
+
+ -
+ Information about job and machine features are now available in
+ glidein startd's classad
+
+
+ -
+ New tool manual_glidein_startup now lets you manually start a
+ glidein for a given factory entry and frontend group
+
+
+ -
+ BOSCO entries can now accept credentials provided from the
+ frontend
+
+
+ -
+ Bug Fix: Glidein is now correctly auto detect RHEL6 and RHEL7
+
+ -
+ Bug Fix: Dagman and Schedd universe jobs are not counted against
+ max jobs running
+
+
+ -
+ Bug Fix: An entry in downtime will now show up in the glidein
+ status
+
+
+ - Bug Fix: Added condor-python rpm as a dependency
+ -
+ Bug Fix: Complex credentials (vm_id+vm_type) are now correctly
+ interpreted
+
+
+ -
+ Bug Fix: Frontend is now more resilent to transient errors
+ communicating with the HTCondor daemons
+
+
+ -
+ Bug Fix: Slave frontend now correctly looks up the master
+ frontend when using condor-python bindings
+
+
+ -
+ Bug Fix: create_condor_tarball now correctly includes required
+ globus libraries from lib/condor that are loaded by HTCondor at
+ runtime
+
+
+ -
+ Bug Fix: Frontend now correctly calculates the ReqMaxGlideins
+ for single core glidein entries
+
+
+ -
+ Bug Fix: Numerical data in glideclient classad is not quoted to
+ preserve the data type
+
+
+ -
+ Bug Fix: Frontend will not request any glideins at entry that is
+ in downtime
+
+
+
+
+ -
+ v3_2_15 released on August 17, 2016 (Manual,
+ Installation instructions,
+ Tarball)
+
+ -
+ Factory will now automatically remove unrecoverable glidein jobs
+ with forcex if they are held for 20 times or more
+
+
+ -
+ Several X509 related attributes from the matching job are now
+ also available in the glidein's startd's classad
+
+
+ -
+ Prefix for attributes created by the periodic scripts is now
+ customizable
+
+
+ -
+ If the glidein detects that a worker node is marked for draining
+ and if it is approaching the drain time, it will now kill the
+ user job
+
+
+ -
+ Frontend can be configured to request idle glideins at all times
+ irrespective of the jobs in the queue
+
+
+ -
+ Bug Fix: Fixed a bug in the frontend downtime command in the
+ script /etc/init.d/gwms-frontend
+
+
+ -
+ Bug Fix: Frontend now correctly considers group credentials
+ before frontend's global credential
+
+
+ -
+ Bug Fix: Installing and upgrading GlideinWMS rpms now correctly
+ trigger httpd and htcondor reload commands in case of EL7
+
+
+ -
+ Bug Fix: Fixed a bug where frontend would crash with pickling
+ error while using htcondor-python bindings
+
+
+ -
+ Bug Fix: Factory now correctly remembers pilots submitted using
+ RFC proxy
+
+
+ -
+ Bug Fix: Fixed a bug where running reconfig or upgrade command
+ on a frontend service would crash
+
+
+
+ -
+ v3_2_14_1 released on June 17, 2016 (Manual,Installation instructions,Tarball)
+
+ - Added support for python in EL7
+ -
+ Updated glidein_startup.sh to make it compatible with factory
+ changes coming in v3_2_15
+
+
+ -
+ Bug Fix: Updated condor_config files used by the Schedd to work
+ with different versions of HTCondor including 8.4.7
+
+
+
+ -
+ v3_2_14 released on June 03, 2016 (Manual,Installation instructions,Tarball)
+
+ -
+ Various curbs and limits triggered in the factory are now logged
+ in the glidefactory and glidefactoryclient classads
+
+
+ - Added initial support for python that comes with EL7
+ -
+ Monitoring stats from factory completed logs are now advertised
+ in the glideresource classads
+
+
+ -
+ Glideins can now shutdown themselves if the worker node is
+ marked for draining with appropriate messages logged in the
+ glideins output
+
+
+ -
+ HTCondor classad fetching is now done by using python bindings
+ by default. Until now this was done using condor_q and
+ condor_status commands
+
+
+ -
+ Various limits configred in the factory and frontend are now
+ advertised in the respective classads
+
+
+ - Updated documentation
+ -
+ Bug Fix: Factory will not release glideins sent to HTCondor CE
+ if they are held with authentication/authorization issues
+
+
+ -
+ Bug Fix: Factory will not release glideins sent to AWS in case
+ of certain types of HoldReason
+
+
+ -
+ Bug Fix: Proxies used by the glidein now if delegated have their
+ lifetime as long as the original proxy
+
+
+ -
+ Bug Fix: Fixed issue where RRD processing was incorrectly
+ throwing ImportError instead of NameError
+
+
+ -
+ Bug Fix: Frontend policies now correctly work if classad
+ attributes like RequestCpus are classad expressions
+
+
+ -
+ Bug Fix: Fixed an issue where a slave frontend in HA mode would
+ crash if the WMS collector was down
+
+
+
+ -
+ v3_2_13 released on March 09, 2016 (Manual,Installation instructions,Tarball)
+
+ - Support XSEDE ProjectId as a credential in frontend
+ -
+ Glidein jobs can now auto detect cpus based on the sites WMS
+
+ -
+ Frontend configuration settings idle_glideins_per_entry,
+ running_glideins_per_entry, running_glideins_total and
+ running_glideins_total_global now consider slots (startd
+ classads) reported in the User collector
+
+
+ -
+ gwms-logcat.sh tool can now forward logs to a folder or
+ http/https url
+
+
+ -
+ Bug Fix: Factory will not release any glideins is max_per_cycle
+ in release section of config is set to 0
+
+
+ -
+ Bug Fix: Frontend now accepts an attr with type="expr" as a
+ condor expression
+
+
+ -
+ Bug Fix: Fixed several issues in the accounting of multi core
+ glideins
+
+
+ -
+ Bug FIx: Counting of idle, running and total jobs in case of
+ multi core glideins is done correctly
+
+
+ -
+ Bug Fix: Frontend will now correctly request enough glideins at
+ sites that support multi core glideins
+
+
+ -
+ Bug Fix: Frontend group limits are now correctly applied based
+ on on the slots rather than glidein (condorg) jobs
+
+
+ -
+ Bug Fix: Fix accounting bug where number of running cores would
+ log a negative count in case of multi core glideins
+
+
+ - Bug Fix: Fixed several issues with the cron type scripts
+
+
+ -
+ v3_2_12_1 released on January 20, 2016 (Manual,Installation instructions,Tarball)
+
+ -
+ Bug Fix: Fixed incompatibility with python 2.4 and bad failure
+ when there is no entry, both introduced in v3_2_12 factory
+ configuration
+
+
+
+ -
+ v3_2_12 released on January 14, 2016 (Manual,Installation instructions,Tarball)
+
+ -
+ Various curbs and limits triggered in the frontend are now
+ logged in the glideresource classads
+
+
+ -
+ Frontend is now more conservative while computing max request
+ running
-
- Bug Fix: create_condor_tarball now correctly includes required
- globus libraries from lib/condor that are loaded by HTCondor at
- runtime
+ Glideins now support advertising custom resources on the worker
+ node This can be used to advertise resources like GPUs. GPUs can
+ also be auto discovered and advertised
-
- Bug Fix: Frontend now correctly calculates the ReqMaxGlideins
- for single core glidein entries
+ Several improvements to rpm packaging. Useful frontend tools are
+ now available in the user path.
-
- Bug Fix: Numerical data in glideclient classad is not quoted to
- preserve the data type
+ Support splitting of factory configuration into factory's
+ deployment specific configuration and entry specific
+ configuration.
-
- Bug Fix: Frontend will not request any glideins at entry that is
- in downtime
+ Support of version control for the factory configuration and
+ splitting of entries configuration from the main factory
+ configuration (entries.d)
+
+
+ -
+ Unique idle jobs matched by the frontend is now available in
+ glideresource classads
+
+
+ -
+ Bug Fix: Fixed a bug where CCB_ADDRESS configuration for the
+ glidein was not created correclty under certain conditions
+
+
+ -
+ Bug Fix: create_frontend script now correctly populates images
+ in the monitoring pages
+
+
+ -
+ Bug Fix: gwms-logcat now correctly supports multiple users
+
+ -
+ Bug Fix: Frontend now correctly deadvertises glideresource
+ classads on shutdown
+
+
+ -
+ Bug Fix: Disable collector's use of shared port to support
+ HTCondor 8.4 (in both factory and user pool)
+
+
+ -
+ Bug Fix: Counting correctly glidein and cores, specially for
+ partitionable slots
+
+
+ -
+ Bug Fix: Fixed bug where DaemonShutdown was failing to consider
+ dynamic slots
+
+
+ -
+ Bug Fix: Fixed bug where NUM_CPUS was not set for partitionable
+ slots resulting in hardware cpu number being used instead of
+ GLIDEIN_CPUS
+
+
+
+ -
+ v3_2_11_2 released on September 18, 2015 (Manual,Installation instructions,Tarball)
+
+ -
+ Fix: Fixed authentication issue introduced in v3_2_11 where a
+ glidein startd fails to send keep alive signals to v8.2.x
+ schedds
+
+
+
+ -
+ v3_2_11_1 released on September 02, 2015 (Manual,Installation instructions,Tarball)
+
+ -
+ Bug Fix: Fixed a bug introduced in v3_2_11 where file period
+ interpreted as number instead of string
+
+
+
+ -
+ v3_2_11 released on August 20, 2015 (Manual,Installation instructions,Tarball)
+
+ -
+ VO Frontend now blacklists schedds with CurbMatchmaking=True
+
+ -
+ You can now over provision Multicore glidein by using
+ GLIDEIN_Resource_Slots attribute to specify different types of
+ resources it provides. For example ioslot
+
+
+ -
+ Glidein can now advertise itself to the site's local HTCondorCE
+ collector if CONDORCE_COLLECTOR_HOST is set in it's environment
+
+
+ -
+ Custom/validation scripts can now be run periodically and not
+ just at the glidein's startup
+
+
+ - Improvements to the rpm packaging
+ - Updated documentation
+ -
+ Bug Fix: Glideins in claimed/idle status are not shutdown by
+ DAEMON_SHUTDOWN expression
+
+
+ - Bug Fix: Fixed a bug in gwms-logcat tool
+ -
+ Bug Fix: The CCB selection behaves correctly and accepts sinful
+ strings
+
+
+ -
+ Bug Fix: Fixed a bug where Frontend under certain conditions
+ would crash because of uninitialized ha_mode
+
+
+ -
+ Bug Fix: Setting GLIDEIN_Report_Failed to ALIVEONLY now creates
+ valid invalidate command
+
+
+ -
+ Bug Fix: Fixed a bug in the accounting of jobs run by a glidein
+ -
+ v3_2_10 released on June 01, 2015 (Manual,Installation instructions,Tarball)
+
+ - Improved Documentation
+ -
+ Bug Fix: Fixed several bugs in accounting of idle and running
+ slots in case of multicore glideins
+
+
+ -
+ Bug Fix: Got rid of old style HTCondor default Memory and Disk
+ requirements in the schedd configuration
+
+ -
+ Bug Fix: DAEMON_SHUTDOWN expression will now let the Multi core
+ glidein to run for appropriate time before killing it
+
+
+
-
- v3_2_15 released on August 17, 2016 (v3_2_9 released on May 08, 2015 (Manual,
- ,Installation instructions,
- ,Tarball)
+ - VO Frontend supports a master-slave HA mode
+ - Added a factory wrapper script to view glidein logs files
+ - Updated the dependency of Glideinwms to HTCondor v8.2.2
+ - Frontend supports CCBs in addition to User Collector
+ - Updated documentation
-
- Factory will now automatically remove unrecoverable glidein jobs
- with forcex if they are held for 20 times or more
+ Bug Fix: glideresource classads now contain appropriate
+ monitoring information
-
- Several X509 related attributes from the matching job are now
- also available in the glidein's startd's classad
+ Bug Fix: Fixed a bug where an unhandled exception would cause a
+ frontend to shutdown
-
- Prefix for attributes created by the periodic scripts is now
- customizable
+ Bug Fix: Removed obsolute defult requirement for vanilla jobs in
+ user schedd's config file
+
+
+ -
+ Bug Fix: Glidein now works correctly when both grid/voms-proxy
+ commands are not available on the worker node
+
+
+ -
+ Bug Fix: GlideinWMSVersion is now correctly reported in rpm
+ distribution
+
+
+
+ -
+ v3_2_8 released on December 30, 2014 (Manual,Installation instructions,Tarball)
+
+ -
+ VO Frontend parameters are added to HTCondor config for ganglia
+ monitoring
+
+
+ -
+ CONDOR_VIEW_HOST is now set to localhost for factory collectors
+ to minimize overhead in communication between primary and
+ secondary collectors
+
+
+ -
+ Added option to compress process logs in factory and frontend
+
+ - Added failed glidein statistics to frontend monitoring
+ -
+ Added idle/running/total core statistics to frontend monitoring
+
+ -
+ Added the support for HTCondor GANGLIAD monitoring (requires
+ HTCondor 8.1 or newer). If you have HTCondor 8.0.x or earlier
+ you must remove
+ /etc/condor/config.d/01_gwms_ganglia.config, otherwise
+ your HTCondor will complain about an unsupported option and
+ crash.
-
- If the glidein detects that a worker node is marked for draining
- and if it is approaching the drain time, it will now kill the
- user job
+ USE_CCB is now enabled by default and this information is
+ advertised in the glidefactory classads
+ - Improved documentation
-
- Frontend can be configured to request idle glideins at all times
- irrespective of the jobs in the queue
+ Bug Fix: Glideins do not mail admins when HTCondor daemon crash
-
-
- Bug Fix: Fixed a bug in the frontend downtime command in the
- script /etc/init.d/gwms-frontend
+ Bug Fix: Gridmanager log paths used by glidein/factory scheds
+ are now correctly expanded for different users
-
- Bug Fix: Frontend now correctly considers group credentials
- before frontend's global credential
+ Bug Fix: Factory and Frontend service start/stop exit codes now
+ confer to Linux standards
-
- Bug Fix: Installing and upgrading GlideinWMS rpms now correctly
- trigger httpd and htcondor reload commands in case of EL7
+ Bug Fix: Fixed issue where work dir and vesioning in frontend
+ config would break the config in case of frontend rpms
-
- Bug Fix: Fixed a bug where frontend would crash with pickling
- error while using htcondor-python bindings
+ Bug Fix: Made improvements to the HTCondor configuration used by
+ factory rpm
-
- Bug Fix: Factory now correctly remembers pilots submitted using
- RFC proxy
+ Bug Fix: There is no name collision for glideins when
+ USE_PID_NAMESPACES is enabled in site's HTCondor batch system
-
- Bug Fix: Fixed a bug where running reconfig or upgrade command
- on a frontend service would crash
+ Bug Fix: Factory does not leak file descriptors when HTCondor
+ commands using privilege separation fail
-
- v3_2_14_1 released on June 17, 2016 (v3_2_7_2 released on November 06, 2014 (Manual,,Installation instructions,,Tarball)
- - Added support for python in EL7
-
- Updated glidein_startup.sh to make it compatible with factory
- changes coming in v3_2_15
+ Bug Fix: Set MASTER.USE_SHARED_PORT instead of USE_SHARED_PORT
+ to avoid secondary collectors using the shared port daemon
-
+
+
+ -
+ v3_2_7_1 released on November 05, 2014 (Manual,Installation instructions,Tarball)
+
-
- Bug Fix: Updated condor_config files used by the Schedd to work
- with different versions of HTCondor including 8.4.7
+ Set USE_SHARED_PORT to get around the issue with HTCondor 8.2.3
-
- v3_2_14 released on June 03, 2016 (v3_2_7 released on October 14, 2014 (Manual,,Installation instructions,,Tarball)
-
- Various curbs and limits triggered in the factory are now logged
- in the glidefactory and glidefactoryclient classads
+ Glideins now have an option to report monitoring info to a
+ different collector configured in the factory
- - Added initial support for python that comes with EL7
+ - Glideins now support shared port
-
- Monitoring stats from factory completed logs are now advertised
- in the glideresource classads
+ Glideins now use local storage for its tmp internal operaations
-
+ - Improved documentation
-
- Glideins can now shutdown themselves if the worker node is
- marked for draining with appropriate messages logged in the
- glideins output
+ Bug Fix: Fixed an issue where a corrupted internal state file
+ would crash the factory
+ - Bug Fix: KeyError in a match_expr is now correctly logged
-
- HTCondor classad fetching is now done by using python bindings
- by default. Until now this was done using condor_q and
- condor_status commands
+ Bug Fix: proxy_url in an entry's config is now correctly used
+
+ -
+ Bug Fix: rrdtool commands are now used correctly when rrdtool
+ python library is not installed on the system
-
- Various limits configred in the factory and frontend are now
- advertised in the respective classads
+ Bug Fix: Error classads now correctly advertise all the relevant
+ attributes
- - Updated documentation
-
- Bug Fix: Factory will not release glideins sent to HTCondor CE
- if they are held with authentication/authorization issues
+ Bug Fix: glidein_off now correctly work with the HTCondor HA
+ setup
+
+ -
+ Bug Fix: Internal security changes are now properly cleaned up
+ and applied in factory and frontend
-
- Bug Fix: Factory will not release glideins sent to AWS in case
- of certain types of HoldReason
+ Bug Fix: Factory monitoring now correctly report UserRunning
+ info when frontend is configured with multiple credentials
-
- Bug Fix: Proxies used by the glidein now if delegated have their
- lifetime as long as the original proxy
+ Bug Fix: Fixed an issue with the factory rpm installation in
+ case of privilege separation
-
- Bug Fix: Fixed issue where RRD processing was incorrectly
- throwing ImportError instead of NameError
+ Bug Fix: Secondary schedd for frontend is now disabled by
+ default
+
+ -
+ Bug Fix: DAEMON_SHUTDOWN in glidein now uses idle timers that
+ are relative to change in the state
+ - Bug Fix: Factory rpm now properly pulls down dependencies
-
- Bug Fix: Frontend policies now correctly work if classad
- attributes like RequestCpus are classad expressions
+ Bug Fix: UpdateSequenceNumber for classads now update correctly
+
+ -
+ Bug Fix: Frontend now correctly provisions multicore glideins if
+ the GLIDEIN_CPUS is configured for the entry
-
- Bug Fix: Fixed an issue where a slave frontend in HA mode would
- crash if the WMS collector was down
+ Bug Fix: GLIDEIN_MaxMemMBs_Estimate now takes GLIDEIN_CPUS in
+ the consideration
-
- v3_2_13 released on March 09, 2016 (v3_2_6 released on July 28, 2014 (Manual,,Installation instructions,,Tarball)
- - Support XSEDE ProjectId as a credential in frontend
- -
- Glidein jobs can now auto detect cpus based on the sites WMS
-
- -
- Frontend configuration settings idle_glideins_per_entry,
- running_glideins_per_entry, running_glideins_total and
- running_glideins_total_global now consider slots (startd
- classads) reported in the User collector
-
-
- -
- gwms-logcat.sh tool can now forward logs to a folder or
- http/https url
-
-
- -
- Bug Fix: Factory will not release any glideins is max_per_cycle
- in release section of config is set to 0
-
-
-
- Bug Fix: Frontend now accepts an attr with type="expr" as a
- condor expression
+ condor_chirp is now added to condor tarbar used by glidein
-
-
- Bug Fix: Fixed several issues in the accounting of multi core
- glideins
+ Added support for submitting glideins to batch sites using
+ BOSCO. Requires HTCondor v8.2.2+
+ - Added new tool to purge old glideins
-
- Bug FIx: Counting of idle, running and total jobs in case of
- multi core glideins is done correctly
+ Added periodic auto-update to Status Now monitoring pages Upon
+ completion, glidein Job history is brought back to the factory
-
- Bug Fix: Frontend will now correctly request enough glideins at
- sites that support multi core glideins
+ Allow for separation of Factory collector and CondorG collector
-
-
- Bug Fix: Frontend group limits are now correctly applied based
- on on the slots rather than glidein (condorg) jobs
+ Bug Fix: Fixed local timezone in some frontend monitoring pages
-
+ - Bug Fix: Improved frontend performance
-
- Bug Fix: Fix accounting bug where number of running cores would
- log a negative count in case of multi core glideins
+ Bug Fix: Requesting single-core partitionable glideins is not
+ allowed
- - Bug Fix: Fixed several issues with the cron type scripts
+ - Bug Fix: Fixed file ownership issues for rpm packages
-
- v3_2_12_1 released on January 20, 2016 (v3_2_5_1 released on June 23, 2014 (Manual,,Installation instructions,,Tarball)
-
- Bug Fix: Fixed incompatibility with python 2.4 and bad failure
- when there is no entry, both introduced in v3_2_12 factory
- configuration
+ Bug Fix: Fixed an issue with the factory_startup template that
+ affects factory reconfig/upgrade in case of RPM
-
- v3_2_12 released on January 14, 2016 (v3_2_5 released on May 19, 2014 (Manual,,Installation instructions,,Tarball)
-
- Various curbs and limits triggered in the frontend are now
- logged in the glideresource classads
-
-
- -
- Frontend is now more conservative while computing max request
- running
-
-
- -
- Glideins now support advertising custom resources on the worker
- node This can be used to advertise resources like GPUs. GPUs can
- also be auto discovered and advertised
-
-
- -
- Several improvements to rpm packaging. Useful frontend tools are
- now available in the user path.
-
-
- -
- Support splitting of factory configuration into factory's
- deployment specific configuration and entry specific
- configuration.
-
-
- -
- Support of version control for the factory configuration and
- splitting of entries configuration from the main factory
- configuration (entries.d)
+ Added administrative commands for frontend fetch_glidein_log,
+ glidein_off and enter_frontend_env
-
- Unique idle jobs matched by the frontend is now available in
- glideresource classads
+ Frontend now considers MAX_JOBS_RUNNING when requesting more
+ glideins
-
- Bug Fix: Fixed a bug where CCB_ADDRESS configuration for the
- glidein was not created correclty under certain conditions
+ Frontend can now perform several tasks in parallel making it
+ more scalable
-
- Bug Fix: create_frontend script now correctly populates images
- in the monitoring pages
+ Frontend and Factory startup scripts are more consistent with
+ each other
+ - Improved Documentation
-
- Bug Fix: gwms-logcat now correctly supports multiple users
+ Bug Fix: Fixed an issue when factory config with HTCondorCE
+ attributes would result in an invalid XML on reconfig
+
-
- Bug Fix: Frontend now correctly deadvertises glideresource
- classads on shutdown
+ Bug Fix: Fixed a bug where number of jobs run as reported by a
+ glidein was significantly scaled up
-
- Bug Fix: Disable collector's use of shared port to support
- HTCondor 8.4 (in both factory and user pool)
+ Bug Fix: Fixed issues in frontend introduced in v3_2_4 where the
+ frontend would crash under certain conditions
-
- Bug Fix: Counting correctly glidein and cores, specially for
- partitionable slots
+ Bug Fix: Frontend's group logging and factory logging now
+ correctly consider the backup_count when configured
-
- Bug Fix: Fixed bug where DaemonShutdown was failing to consider
- dynamic slots
+ Bug Fix: Frontend RPM now corectly creates frontend.xml config
+ with default trust_domain='grid' for the credentials
-
- Bug Fix: Fixed bug where NUM_CPUS was not set for partitionable
- slots resulting in hardware cpu number being used instead of
- GLIDEIN_CPUS
+ Bug Fix: Fixed frontend performance issue introduced in v3.2.4
-
- v3_2_11_2 released on September 18, 2015 (v3_2_4 released on April 14, 2014 (Manual,,Installation instructions,,Tarball)
+ - Added support for HTCondor-CE attributes in the factory
-
- Fix: Fixed authentication issue introduced in v3_2_11 where a
- glidein startd fails to send keep alive signals to v8.2.x
- schedds
+ Made several performance improvements to frontend. Frontend does
+ several tasks in parallel to better utilize the CPU.
-
-
- -
- v3_2_11_1 released on September 02, 2015 (Manual,Installation instructions,Tarball)
-
+
-
- Bug Fix: Fixed a bug introduced in v3_2_11 where file period
- interpreted as number instead of string
+ Factory and frontend monitoring pages now use new javascriptrrd
+ v1.1.0+
+
+
+ - Factory monitoring now aggregates Log RRDs
+ - Frontend can now limit total idle glideins
+ - Added limits to globaly total idle glideins
+ - Added badput summary line in the factory report
+ - Improved documentation
+ -
+ Bug Fix: Factory and frontend operations like reconfig and
+ upgrade now check if they are run by valid users
+
+
+ - Bug Fix: Fixed partitioning of multi-core glideins
+ - Bug Fix: Fixed bug in factory/frontend stopping
+ -
+ Bug Fix: Fixed several bugs in the /etc/init.d/gwms-factory
+ script available through the RPM distribution
+
+
+ -
+ Bug Fix: Fixed bug with the factory/frontend monitoring that
+ resulted in significantly scaled up monitoring numbers when
+ frontend used multiple proxies
+
+
+ -
+ Bug Fix: Factory now properly advertises entries in downtime
+
+ -
+ Bug Fix: GLIDEIN_Glexec_Use when defined in the group now
+ correctly overriddes the value defined in global scope
-
- v3_2_11 released on August 20, 2015 (v3_2_3 released on February 03, 2014 (Manual,,Installation instructions,,Tarball)
-
- VO Frontend now blacklists schedds with CurbMatchmaking=True
+ Glideins now have the ability to track the worker node batch
+ slot. Based on the batch system at site (HTCondor, SGE, PBS,
+ LSF, SLURM) the information is reported in the glidein's STARTD
+ classad and logged in job's log file written by HTCondor using
+ classad variables GLIDEIN_SiteWMS, GLIDEIN_SiteWMS_JOBID,
+ GLIDEIN_SiteWMS_QUEUE and GLIDEIN_SiteWMS_SLOT.
+
-
- You can now over provision Multicore glidein by using
- GLIDEIN_Resource_Slots attribute to specify different types of
- resources it provides. For example ioslot
+ Number of rotated process_logs for factory and frontend process
+ to keep can now be configured using backup_count configuration
+ attribute
-
- Glidein can now advertise itself to the site's local HTCondorCE
- collector if CONDORCE_COLLECTOR_HOST is set in it's environment
+ Bug Fix: Factory now correctly updates the renewed credentials
+ it gets from the Frontends
-
- Custom/validation scripts can now be run periodically and not
- just at the glidein's startup
+ Bug Fix: Cloud related configuration attributes
+ VM_DISABLE_SHUTDOWN and VM_MAX_LIFETIME are now documented
- - Improvements to the rpm packaging
- - Updated documentation
-
- Bug Fix: Glideins in claimed/idle status are not shutdown by
- DAEMON_SHUTDOWN expression
+ Bug Fix: Partitionable slots now correctly evaluate daemon
+ shutdown
-
- - Bug Fix: Fixed a bug in gwms-logcat tool
-
- Bug Fix: The CCB selection behaves correctly and accepts sinful
- strings
+ Bug Fix: Partitionable slots now correctly coalesce when the
+ jobs finish
-
- Bug Fix: Fixed a bug where Frontend under certain conditions
- would crash because of uninitialized ha_mode
+ Bug Fix: For non-rpm installs, reconfiguring the factory from
+ outside the factory working directory now works correctly
-
- Bug Fix: Setting GLIDEIN_Report_Failed to ALIVEONLY now creates
- valid invalidate command
+ Bug Fix: Frontend with no credentials configured logs
+ appropriate info in the log files
-
- Bug Fix: Fixed a bug in the accounting of jobs run by a glidein
+ Bug Fix: clone_glidein tool is now packaged with the factory rpm
+
+ -
+ Bug Fix: Factory now correctly cleans up completed_jobs logs
+
+ -
+ Bug Fix: Improved Frontend efficiency by reducing the calls to
+ OpenSSL
-
- v3_2_10 released on June 01, 2015 (v3_2_2 released on November 08, 2013 (Manual,,Installation instructions,,Tarball)
- - Improved Documentation
- -
- Bug Fix: Fixed several bugs in accounting of idle and running
- slots in case of multicore glideins
-
-
-
- Bug Fix: Got rid of old style HTCondor default Memory and Disk
- requirements in the schedd configuration
+ Bug Fix: Fix a bug where factory would crash if it fails to
+ query client global classads in wms collector
-
- Bug Fix: DAEMON_SHUTDOWN expression will now let the Multi core
- glidein to run for appropriate time before killing it
+ Bug Fix: Glidein jobs correctly interpret the ARC CE FINISHED
+ state
-
- v3_2_9 released on May 08, 2015 (v3_2_1 released on October 30, 2013 (Manual,,Installation instructions,,Tarball)
- - VO Frontend supports a master-slave HA mode
- - Added a factory wrapper script to view glidein logs files
- - Updated the dependency of Glideinwms to HTCondor v8.2.2
- - Frontend supports CCBs in addition to User Collector
- - Updated documentation
-
- Bug Fix: glideresource classads now contain appropriate
- monitoring information
+ v3_2_1 Added support for a plug-in architecture for config that
+ lets admins manipulate frontend and factory config with ease
-
- Bug Fix: Fixed a bug where an unhandled exception would cause a
- frontend to shutdown
+ Bug Fix: Factory accounting now correctly accounts for held jobs
-
-
- Bug Fix: Removed obsolute defult requirement for vanilla jobs in
- user schedd's config file
+ Bug Fix: Improved error reporting in case of misconfigured
+ credentials
-
- Bug Fix: Glidein now works correctly when both grid/voms-proxy
- commands are not available on the worker node
+ Bug Fix: Improved error reporting when factory fails to startup
-
+ - Bug Fix: Improved factory performance during log cleanup
-
- Bug Fix: GlideinWMSVersion is now correctly reported in rpm
- distribution
+ Bug Fix: Fixed a bug introduced in v3_2 where glidein update
+ interval was too short that resulted in performance issues for
+ busy collectors
-
- v3_2_8 released on December 30, 2014 (v3_2 released on October 10, 2013 (Manual,,Installation instructions,,Tarball)
-
- VO Frontend parameters are added to HTCondor config for ganglia
- monitoring
+ Added support for generation of condor tarball from condor
+ installed via rpm
+
+ -
+ Bug Fix: Fixed a race condition while shutting down factory
+ service
+
+
+ -
+ Bug Fix: Fixed a bug where glidein would not set certain condor
+ config variables correctly
+
+
+ -
+ Bug Fix: Fixed a bug where analyze_queues and analyze_entries
+ would not work with the http:// URI
+
+
+ -
+ Bug Fix: Collector port ranges in frontend.xml are now validated
+ for errors
+
+
+ -
+ Bug Fix: Schedd name, frontend name and identity in factory
+ config are validated for errors
+
+
+ -
+ Bug Fix: Starting a factory with all entries disabled now prints
+ helpful message
+
+
+ -
+ Bug Fix: Glidein does not leak LD_LIBRARY_PATH to job's
+ environment
+
+
+ -
+ Bug Fix: Added several speed enhancements to the factory to get
+ looptime under acceptable limits
-
- CONDOR_VIEW_HOST is now set to localhost for factory collectors
- to minimize overhead in communication between primary and
- secondary collectors
+ Bug Fix: Fixed a bug where factory would not query rpm installed
+ schedd correctly
-
- Added option to compress process logs in factory and frontend
+ Bug Fix: Fixed a bug where factory would throw exception while
+ logging during aggregating monitoring information
- - Added failed glidein statistics to frontend monitoring
+
-
- Added idle/running/total core statistics to frontend monitoring
+ Bug Fix: Fixed a bug where factory would fail advertising
+ classads to the WMS Collector when classad attributes had quotes
+ in them
+
-
- Added the support for HTCondor GANGLIAD monitoring (requires
- HTCondor 8.1 or newer). If you have HTCondor 8.0.x or earlier
- you must remove
- /etc/condor/config.d/01_gwms_ganglia.config, otherwise
- your HTCondor will complain about an unsupported option and
- crash.
+ Bug Fix: Fixed a bug where glidein's condor_startd would not
+ correctly advertise LSB_DESCRIPTION in the classad correctly
-
- USE_CCB is now enabled by default and this information is
- advertised in the glidefactory classads
+ Bug Fix: Fixed a bug where frontend would not correctly account
+ idle jobs that used vanilla proxies
- - Improved documentation
-
- Bug Fix: Glideins do not mail admins when HTCondor daemon crash
+ Bug Fix: Factory now properly logs errors when authenticating a
+ frontend
+
-
- Bug Fix: Gridmanager log paths used by glidein/factory scheds
- are now correctly expanded for different users
+ Bug Fix: Fixed a bug in glidein where it would not report back
+ to all the user pool collectors in HA mode correctly
-
- Bug Fix: Factory and Frontend service start/stop exit codes now
- confer to Linux standards
+ Bug Fix: Fixed a bug where some of the parameters passed to
+ glideins were not escaped correctly
-
- Bug Fix: Fixed issue where work dir and vesioning in frontend
- config would break the config in case of frontend rpms
+ Bug Fix: Submitting glideins in test only mode now works
+ correctly
+
+ -
+ Bug Fix: Fixed a bug where factory would not submit glideins
+ when privilege separation is disabled
-
- Bug Fix: Made improvements to the HTCondor configuration used by
- factory rpm
+ Bug Fix: Fixed a bug in manageFactoryDowntimes.py where it could
+ not find required python libraries
-
- Bug Fix: There is no name collision for glideins when
- USE_PID_NAMESPACES is enabled in site's HTCondor batch system
+ Bug Fix: Fixed a bug in factory accounting where glideins in
+ certain state were not accounted towards idle state
+ - Bug Fix: Added support for the ACCEPTED state for ARC CE
-
- Bug Fix: Factory does not leak file descriptors when HTCondor
- commands using privilege separation fail
+ Bug Fix: factoryStatus.html now correctly auto-selects timezone
-
-
- -
- v3_2_7_2 released on November 06, 2014 (Manual,Installation instructions,Tarball)
-
-
- Bug Fix: Set MASTER.USE_SHARED_PORT instead of USE_SHARED_PORT
- to avoid secondary collectors using the shared port daemon
+ Bug Fix: Fixed a bug where RSL for NorduGrid CE was not
+ populated
-
-
- -
- v3_2_7_1 released on November 05, 2014 (Manual,Installation instructions,Tarball)
-
-
- Set USE_SHARED_PORT to get around the issue with HTCondor 8.2.3
+ Bug Fix: glidefactory classads now correctly display
+ GlideinWMSVersion
+
+
+
+
+ Development Series
+ This is work in progress, so some pieces may be not so tested yet and
+ the documentation out of sync. But it passed all out tests and is a good
+ place to look for what is going on.
+
+
+
-
- v3_2_7 released on October 14, 2014 (v3_9_6 released on October 27, 2022 (Manual,,
+ Installation instructions,Tarball)
-
- Glideins now have an option to report monitoring info to a
- different collector configured in the factory
+ Added token authentication to Glideins running in the Cloud (AWS
+ and GCE). Now Glideins on all supported resources can
+ authenticate back using IDTOKENS when using recent HTCSS
+ versions.
-
- - Glideins now support shared port
-
- Glideins now use local storage for its tmp internal operaations
+ Added `GLIDEIN_PERIODIC_SCRIPT` env variable to notify periodic
+ custom scripts
- - Improved documentation
-
- Bug Fix: Fixed an issue where a corrupted internal state file
- would crash the factory
+ Added the possibility to set the Glidein HTCSS TRUST_DOMAIN as
+ attribute in the Frontend configuration
-
- - Bug Fix: KeyError in a match_expr is now correctly logged
-
- Bug Fix: proxy_url in an entry's config is now correctly used
+ The Frontend configuration is now valid (reconfig/upgrade
+ successful) even if some HTCSS schedds are not in DNS. Failing
+ only if all schedds are unknown to DNS
-
- Bug Fix: rrdtool commands are now used correctly when rrdtool
- python library is not installed on the system
+ Bug fix: Fixed `glidien_config` corrupted by concurrent custom
+ scripts run via HTCSS startd cron (#163)
-
-
- Bug Fix: Error classads now correctly advertise all the relevant
- attributes
+ Bug fix: Fixed `setup_x509.sh` not to write to stdout when
+ running as periodic script in HTCSS start cron (issues#162,#164)
-
-
- Bug Fix: glidein_off now correctly work with the HTCondor HA
- setup
+ Bug fix: Fixed setup_x509.sh creates proxy file in directory
+ used for tokens (issue#201)
-
- Bug Fix: Internal security changes are now properly cleaned up
- and applied in factory and frontend
+ Bug fix: Fixed GLIDEIN_START_DIR_ORIG and GLIDEIN_WORKSPACE_ORIG
+ values in glidein_config
-
-
- Bug Fix: Factory monitoring now correctly report UserRunning
- info when frontend is configured with multiple credentials
+ Bug fix: Fixed unnecessary proxy/hostcert.pem workaround in
+ frontend config (issue#66)
-
-
- Bug Fix: Fixed an issue with the factory rpm installation in
- case of privilege separation
+ Bug fix: Fixed analyze_entries and python3 readiness (issue#194)
-
-
- Bug Fix: Secondary schedd for frontend is now disabled by
- default
+ Bug fix: Fixed gwms-renew-proxies service should check if local
+ VOMS cert is expired (issue#21)
-
- Bug Fix: DAEMON_SHUTDOWN in glidein now uses idle timers that
- are relative to change in the state
+ Bug fix: Fixed python3 check return value in case of exception
+ (PR#211)
-
- - Bug Fix: Factory rpm now properly pulls down dependencies
-
- Bug Fix: UpdateSequenceNumber for classads now update correctly
+ Bug fix: Fixed list_get_intersection in singularity_lib.sh that
+ was requiring python2 (PR#212)
-
- Bug Fix: Frontend now correctly provisions multicore glideins if
- the GLIDEIN_CPUS is configured for the entry
+ Bug fix: Unset SEC_PASSWORD_DIRECTORY in the Glidein HTCSS
+ configuration, was causing warnings for unknown files (PR#226)
-
-
- Bug Fix: GLIDEIN_MaxMemMBs_Estimate now takes GLIDEIN_CPUS in
- the consideration
+ Bug fix: HTCSS DC_DAEMON_LIST now is equal to DAEMON_LIST only
+ in the Factory, in all other GlideinWMS components only selected
+ HTCSS daemons are added explicitly to it (issue#205)
+
+ -
+ Bug fix: Working and local tmp directories are removed during
+ Glidein cleanup also when the start directory is missing. This
+ result in a loss of Glidein final status information but avoids
+ sandbox leaks on the Worker Node. (issue#189)
+
+ -
+ NOTE: custom scripts should always read values via
+ gconfig_get(). The only exception is the parsing or the line to
+ get the add_config_line source file
+
+ -
+ NOTE: add_config_line is deprecated in favor of gconfig_add.
+ Will be removed from future versions
+
+ -
+ NOTE: add_config_line_safe is deprecated in favor of
+ gconfig_add_safe. gconfig_add is the recommended method to use
+ also in concurrent scripts
+
+ -
+ NOTE: gconfig_add_unsafe, with a similar behavior to the former
+ add_config_line, has been added. Its use is not recommended,
+ unless absolutely sure that no script will be run in parallel
+
+ -
+ NOTE: Only the first collector in TRUST_DOMAIN is kept,
+ following collectors are removed. This happens both in the
+ Frontend token issuer and in the setting of the Glidein
+ TRUST_DOMAIN (setup_x509.sh).
-
- v3_2_6 released on July 28, 2014 (v3_9_5 released on May 17, 2022 (Manual,,
+ Installation instructions,Tarball)
+ - Support for credential generator plugins
+ - Support for Apptainer
-
- condor_chirp is now added to condor tarbar used by glidein
+ Support for python3-rrdtool added in OSG 3.6 also for RHEL7
-
- Added support for submitting glideins to batch sites using
- BOSCO. Requires HTCondor v8.2.2+
+ Support for CVMFS on-demand via cvmfsexec (mount/umountrepo)
-
- - Added new tool to purge old glideins
+ - Configurable idtokens lifetime
+ - Improved frontend logging
+ - Improved default SHARED_PORT configuration
+ - Special handling of multiline condor config values
+ - Advertise CONTAINERSW_ variables
-
- Added periodic auto-update to Status Now monitoring pages Upon
- completion, glidein Job history is brought back to the factory
+ Bug fix: Fixes an issue causing SciTokens transfered to the
+ factory to be corrupted
-
-
- Allow for separation of Factory collector and CondorG collector
+ Bug fix: Fixes to manual_glidein_submit for compatibility with
+ Python 3
-
- Bug Fix: Fixed local timezone in some frontend monitoring pages
+ Bug fix: Fixes to OSG_autoconf for compatibility with Python 3
- - Bug Fix: Improved frontend performance
+ - Bug fix: Fixes a minor issue in job.condor
+ - Bug fix: Fixes OS detection for CentOS Stream 8
-
- Bug Fix: Requesting single-core partitionable glideins is not
- allowed
+ Bug fix: Fixes an issue with scitoken handling to gsi_proxy
+ configured factory entries
+
+ - Bug fix: Fixes condor tarball directory structure
+ -
+ Bug fix: Stops clearing LD_LIBRARY_PATH before starting HTCondor
+
+ -
+ Bug fix: Ensures that idtokens are transferred to glidein
+ credential directory
-
- - Bug Fix: Fixed file ownership issues for rpm packages
-
- v3_2_5_1 released on June 23, 2014 (v3_9_4 released on Jan 25, 2022 (Manual,,
+ Installation instructions,Tarball)
+ - Includes all features and fixes of 3.7.6
+ - Adds support for CentOS 8
-
- Bug Fix: Fixed an issue with the factory_startup template that
- affects factory reconfig/upgrade in case of RPM
+ Adds two new RPM packages (vofrontend-libs and
+ vofrontend-glidein) to provide essential libraries for
+ frontend-like applications
+
+ - Updates license to Apache 2.0
+ -
+ Bug fix: Updates libraries for compatibility with PyYAML 6.0
+
+ -
+ Bug fix: Fixes a sring encoding issue affecting the cat log
+ scripts
+ - Bug fix: Fixes AWS/Google metadata encoding error
-
- v3_2_5 released on May 19, 2014 (v3_9_3 released on Sep 21, 2021 (Manual,,
+ Installation instructions,Tarball)
+ - Includes all features and fixes of 3.7.5
+ - Adds type validation to the check_python3_expr.py script
+ - Removes unused encodingSupport.py and its unit tests
-
- Added administrative commands for frontend fetch_glidein_log,
- glidein_off and enter_frontend_env
+ Bug fix: Fixes an encoding issue affecting cloud submissions
-
+
+
+ -
+ v3_9_2 released on Jun 1, 2021 (Manual,
+ Installation instructions)
+
+ - Includes all features and fixes of 3.7.3 and 3.7.4
+ - Drops support for BDII along with its LDAP requirements
-
- Frontend now considers MAX_JOBS_RUNNING when requesting more
- glideins
+ Adds check_python3_expr.py to validate python3 in match
+ expressions
-
+ - Bug fix: Fixes string encoding issues on glideFactoryLib
-
- Frontend can now perform several tasks in parallel making it
- more scalable
+ Bug fix: Fixes an issue that would cause some factory
+ configurations to generate malformed Condor files
-
+
+
+ -
+ v3_9_1 released on February 11, 2021 (Manual,
+ Installation instructions)
+
+ - Includes all features and fixes of 3.7.2
+ - Separates httpd packages from Factory and Frontend RPMs
-
- Frontend and Factory startup scripts are more consistent with
- each other
+ Bug fix: Frontend no longer reports duplicate glideclient
+ daemons to the Factory
-
- - Improved Documentation
+
+
+ -
+ v3_9 released on September 18, 2020 (Manual,
+ Installation instructions)
+
+ - Includes all features and fixes of 3.6.3 and 3.7.1
-
- Bug Fix: Fixed an issue when factory config with HTCondorCE
- attributes would result in an invalid XML on reconfig
+ Python 3 code, some pitfalls to pay attention to are in the Bug
+ Fix section
-
-
- Bug Fix: Fixed a bug where number of jobs run as reported by a
- glidein was significantly scaled up
+ Bug fix: custom disctionaries with __getattr__ in the creation
+ package (an infinite loop was masked in Python 2)
-
+ - Bug fix: str vs bytes, especially in crypto libraries
-
- Bug Fix: Fixed issues in frontend introduced in v3_2_4 where the
- frontend would crash under certain conditions
+ Bug fix: None cannot be used in sorted string lists and we use
+ (None, None, None) for unmatched jobs in the Frontend
-
+ - Bug fix: Add __repr__ method to SubParams class
-
- Bug Fix: Frontend's group logging and factory logging now
- correctly consider the backup_count when configured
+ Bug fix: Fixed pycache of creation/lib included in multiple
+ packages
-
-
- Bug Fix: Frontend RPM now corectly creates frontend.xml config
- with default trust_domain='grid' for the credentials
+ Bug fix: Fixed spec file to include python3 compiled files and
+ not python2 ones
-
+ - Bug fix: Fixed Mapping import
-
- Bug Fix: Fixed frontend performance issue introduced in v3.2.4
+ Bug fix: Fixed str vs bytes confusion in cryptography functions
+ and classes
-
- v3_2_4 released on April 14, 2014 (v3_7_7 released on June 21 2022 (Manual,,
+ Installation instructions,Tarball)
- - Added support for HTCondor-CE attributes in the factory
- -
- Made several performance improvements to frontend. Frontend does
- several tasks in parallel to better utilize the CPU.
-
-
- -
- Factory and frontend monitoring pages now use new javascriptrrd
- v1.1.0+
-
-
- - Factory monitoring now aggregates Log RRDs
- - Frontend can now limit total idle glideins
- - Added limits to globaly total idle glideins
- - Added badput summary line in the factory report
- - Improved documentation
+ - Authenticate w/ tokens w/ the HTCondor-CE collector
-
- Bug Fix: Factory and frontend operations like reconfig and
- upgrade now check if they are run by valid users
-
-
- - Bug Fix: Fixed partitioning of multi-core glideins
- - Bug Fix: Fixed bug in factory/frontend stopping
- -
- Bug Fix: Fixed several bugs in the /etc/init.d/gwms-factory
- script available through the RPM distribution
+ Added CONTINUE_IF_NO_PROXY attribute to control GSI to token
+ transition
-
-
- Bug Fix: Fixed bug with the factory/frontend monitoring that
- resulted in significantly scaled up monitoring numbers when
- frontend used multiple proxies
+ Bug fix: Fixes an issue causing SciTokens transfered to the
+ factory to be corrupted
-
-
- Bug Fix: Factory now properly advertises entries in downtime
+ Bug fix: Fixes condor_config -dump is not safe if the pool
+ config has multiline entries
+ - Bug fix: Add job.condor typo fix for bosco to 3.7.7
-
- Bug Fix: GLIDEIN_Glexec_Use when defined in the group now
- correctly overriddes the value defined in global scope
+ NOTE: Now branch_v3_7 is completely separate. Feature or fixes
+ have to be backported
-
- v3_2_3 released on February 03, 2014 (v3_7_6 released on Jan 18 2022 (Manual,,
+ Installation instructions,Tarball)
+ - Flexible mount point for CVMFS in glidein
+ - Frontend and factory check expiration of SciTokens
+ - Shared port is default for User Collector
+ - API to retrieve job information from condor jobID
+ - OSG_autoconf now allows override yaml configuration files
-
- Glideins now have the ability to track the worker node batch
- slot. Based on the batch system at site (HTCondor, SGE, PBS,
- LSF, SLURM) the information is reported in the glidein's STARTD
- classad and logged in job's log file written by HTCondor using
- classad variables GLIDEIN_SiteWMS, GLIDEIN_SiteWMS_JOBID,
- GLIDEIN_SiteWMS_QUEUE and GLIDEIN_SiteWMS_SLOT.
+ expiration and other validity of JWT tokens checked before use
-
-
- Number of rotated process_logs for factory and frontend process
- to keep can now be configured using backup_count configuration
- attribute
+ Bug Fix: condor_chirp no longer fail when python2 used in job
-
-
- Bug Fix: Factory now correctly updates the renewed credentials
- it gets from the Frontends
+ Bug Fix: IDTOKEN issuer changed from collector host to trust
+ domain
-
-
- Bug Fix: Cloud related configuration attributes
- VM_DISABLE_SHUTDOWN and VM_MAX_LIFETIME are now documented
+ Bug Fix: proxy was failing to renew when both SCITOKEN and
+ grid_proxy auth used
-
-
- Bug Fix: Partitionable slots now correctly evaluate daemon
- shutdown
+ Bug Fix: OSG_autoconf automatic assignment of submit attributes
+ fixed
+ - Bug Fix: OSG_autoconf common entry fields implemented
-
- Bug Fix: Partitionable slots now correctly coalesce when the
- jobs finish
+ Bug Fix: OSG_autoconf now possible to append instead of replace
+ values
-
-
- Bug Fix: For non-rpm installs, reconfiguring the factory from
- outside the factory working directory now works correctly
+ Bug Fix: glidein clears LD_LIBRARY_PATH prior to starting
+ HTCondor
-
-
- Bug Fix: Frontend with no credentials configured logs
- appropriate info in the log files
+ Bug Fix: JOB_TOKENS no longer set in glidein environment if not
+ needed
-
+ - Bug Fix: IDTOKEN/SCITOKEN auth to bosco sites fixed
+ - Bug Fix: improved IDTOKEN revocation procedures
+ - Bug Fix: SSL added to condor *_AUTHENTICATION_METHODS
-
- Bug Fix: clone_glidein tool is now packaged with the factory rpm
+ Bug Fix: Possible to have frontend group with only SCITOKEN
+ authentication
-
- Bug Fix: Factory now correctly cleans up completed_jobs logs
+ Bug Fix: Factory stores frontend group information per-token
-
- Bug Fix: Improved Frontend efficiency by reducing the calls to
- OpenSSL
+ Bug Fix: Intermittent -check_v35 error fails to startup factory
+
+ -
+ Bug Fix: Fix glidein_starup.sh in staging area to be one with
+ payload
+
+ - Bug Fix: GCB support removed
+ -
+ Bug Fix: glidein_cleanup can cleanup directories chmoded
+ read-only
+ - Bug Fix: improved shellcheck testing
+ - Bug Fix: corrected paths for some variables in glideins
+ - Bug Fix: eliminated spurious log messages in glideins
+ - Bug Fix: STARTD_JOB_ATTRS fixed in glideins
+ - Bug Fix: GWMS_PATH in glidein_startup.sh corrected
+ - Bug Fix: better cleanup of tmpfiles after install
-
- v3_2_2 released on November 08, 2013 (v3_7_5 released on Sep 2 2021 (Manual,,
+ Installation instructions,Tarball)
+ - Support for big files in the glideinwms repository
+ - Additional download mechanism for files
+ - Bug Fix: Have SciToken credential be per-entry
+ - Bug Fix: Fix IDTOKEN generation in the Frontend
-
- Bug Fix: Fix a bug where factory would crash if it fails to
- query client global classads in wms collector
+ Bug Fix: Cleanup script not executing correctly at glidein
+ termination
-
-
- Bug Fix: Glidein jobs correctly interpret the ARC CE FINISHED
- state
+ Bug Fix: Automount of CVMFS integrated into glidein script
-
- v3_2_1 released on October 30, 2013 (v3_7_4 released on May 10, 2021 (Manual,,
+ Installation instructions,Tarball)
-
- v3_2_1 Added support for a plug-in architecture for config that
- lets admins manipulate frontend and factory config with ease
-
-
- -
- Bug Fix: Factory accounting now correctly accounts for held jobs
-
- -
- Bug Fix: Improved error reporting in case of misconfigured
- credentials
-
-
- -
- Bug Fix: Improved error reporting when factory fails to startup
+ Consider not available (not Idle) partitionable slots where all
+ GPUs are used
- - Bug Fix: Improved factory performance during log cleanup
-
- Bug Fix: Fixed a bug introduced in v3_2 where glidein update
- interval was too short that resulted in performance issues for
- busy collectors
+ Bug Fix: Submissions from some 3.6.5 frontends to 3.7.3
+ factories go on hold
-
- v3_2 released on October 10, 2013 (v3_7_3 released on March 25, 2021 (Manual,,
+ Installation instructions,Tarball)
+ - More efficient generation of IDTOKENS by frontend
-
- Added support for generation of condor tarball from condor
- installed via rpm
-
- -
- Bug Fix: Fixed a race condition while shutting down factory
- service
-
-
- -
- Bug Fix: Fixed a bug where glidein would not set certain condor
- config variables correctly
+ SciToken credentials configured and treated same as other
+ credentials for glidein submission
-
-
- Bug Fix: Fixed a bug where analyze_queues and analyze_entries
- would not work with the http:// URI
+ Glideins configured without GSI but with SciToken and IDTOKENS
+ successfully run jobs.
-
+ - Order of places glidein checks for singularity changed
-
- Bug Fix: Collector port ranges in frontend.xml are now validated
- for errors
+ PATH with multiple directories can be configured to override
+ singularity binary
-
+ - Remove LD_PRELOAD from glidein environment by default
-
- Bug Fix: Schedd name, frontend name and identity in factory
- config are validated for errors
+ Fail without fall-back when user provided singularity image
+ missing
-
+ - Improved GPU implementation in glidenins
-
- Bug Fix: Starting a factory with all entries disabled now prints
- helpful message
+ OSG_autoconf more configurable, auth_method and pilot_entries
+ can be set
-
-
- Bug Fix: Glidein does not leak LD_LIBRARY_PATH to job's
- environment
+ Bug Fix: Missing GSI configuration does not stop schedd query
+ when IDTOKENS configured.
-
-
- Bug Fix: Added several speed enhancements to the factory to get
- looptime under acceptable limits
+ Bug Fix: Factory should not check condor tarball version prior
+ to SciToken authentication to CE
-
-
- Bug Fix: Fixed a bug where factory would not query rpm installed
- schedd correctly
+ Bug Fix: tmpfile needs to be on correct file system when factory
+ writes tokens from frontend
-
-
- Bug Fix: Fixed a bug where factory would throw exception while
- logging during aggregating monitoring information
+ Bug Fix: IDTOKENS auth does not fail on improper condor_version
+ comparison
-
+ - Bug Fix: rpm sub package (http) building issues fixed
+
+
+ -
+ v3_7_2 released on December 21, 2020 (Manual,
+ Installation instructions)
+
+ - Includes all features and fixes of 3.6.6
+ - Drop GLExec support
-
- Bug Fix: Fixed a bug where factory would fail advertising
- classads to the WMS Collector when classad attributes had quotes
- in them
+ Mechanism for running custom scripts at the end of glidein
+ execution
-
-
- Bug Fix: Fixed a bug where glidein's condor_startd would not
- correctly advertise LSB_DESCRIPTION in the classad correctly
+ OSG_autoconf protection added to prevent failures when CE
+ dissapears from OSG Collector
-
-
- Bug Fix: Fixed a bug where frontend would not correctly account
- idle jobs that used vanilla proxies
+ Bug fix: Frontend no longer over-generates tokens for entries
-
-
- Bug Fix: Factory now properly logs errors when authenticating a
- frontend
+ Bug fix: Factory does not remove glideins when idle limit is hi
-
+
+
+ -
+ v3_7_1 released on November 3, 2020 (Manual,
+ Installation instructions)
+
+ - Includes all features and fixes of 3.6.5
-
- Bug Fix: Fixed a bug in glidein where it would not report back
- to all the user pool collectors in HA mode correctly
+ Configuration changes to make more compatible with HTCondor 8.9
+ on initial install
-
+ - SciTokens authentication between Factory and CE
+ - IDTokens authentication between Factory and Frontend
-
- Bug Fix: Fixed a bug where some of the parameters passed to
- glideins were not escaped correctly
+ Bug fix: Factory reconfigs and startups do not require manual
+ creation of directories
-
-
- Bug Fix: Submitting glideins in test only mode now works
+ Bug fix: Factory builds Condor Tarballs to send with glideins
correctly
+
+
+ -
+ v3_7 released on March 31, 2020 (Manual,
+ Installation instructions,Tarball)
+
+ - Includes all features and fixes of 3.6.2
+ - Use of HTCondor token-auth for Glideins authentication
+ - Added secure logging channel
-
- Bug Fix: Fixed a bug where factory would not submit glideins
- when privilege separation is disabled
-
-
- -
- Bug Fix: Fixed a bug in manageFactoryDowntimes.py where it could
- not find required python libraries
-
-
- -
- Bug Fix: Fixed a bug in factory accounting where glideins in
- certain state were not accounted towards idle state
-
-
- - Bug Fix: Added support for the ACCEPTED state for ARC CE
- -
- Bug Fix: factoryStatus.html now correctly auto-selects timezone
-
- -
- Bug Fix: Fixed a bug where RSL for NorduGrid CE was not
- populated
-
- -
- Bug Fix: glidefactory classads now correctly display
- GlideinWMSVersion
+ Refactored glidien_startup.sh to separate out the code in
+ heredoc sections and improve code quality
-
-
-
-
- Development Series
- This is work in progress, so some pieces may be not so tested yet and
- the documentation out of sync. But it passed all out tests and is a good
- place to look for what is going on.
-
-
-
--
- v3_9_6 released on October 27, 2022 (Manual,
- Installation instructions)
-
- - Added token authentication to Glideins running in the Cloud (AWS and GCE). Now Glideins on all supported resources can authenticate back using IDTOKENS when using recent HTCSS versions.
- - Added `GLIDEIN_PERIODIC_SCRIPT` env variable to notify periodic custom scripts
- - Added the possibility to set the Glidein HTCSS TRUST_DOMAIN as attribute in the Frontend configuration
- - The Frontend configuration is now valid (reconfig/upgrade successful) even if some HTCSS schedds are not in DNS. Failing only if all schedds are unknown to DNS
- - Bug fix: Fixed `glidien_config` corrupted by concurrent custom scripts run via HTCSS startd cron (#163)
- - Bug fix: Fixed `setup_x509.sh` not to write to stdout when running as periodic script in HTCSS start cron (issues#162,#164)
- - Bug fix: Fixed setup_x509.sh creates proxy file in directory used for tokens (issue#201)
- - Bug fix: Fixed GLIDEIN_START_DIR_ORIG and GLIDEIN_WORKSPACE_ORIG values in glidein_config
- - Bug fix: Fixed unnecessary proxy/hostcert.pem workaround in frontend config (issue#66)
- - Bug fix: Fixed analyze_entries and python3 readiness (issue#194)
- - Bug fix: Fixed gwms-renew-proxies service should check if local VOMS cert is expired (issue#21)
- - Bug fix: Fixed python3 check return value in case of exception (PR#211)
- - Bug fix: Fixed list_get_intersection in singularity_lib.sh that was requiring python2 (PR#212)
- - Bug fix: Unset SEC_PASSWORD_DIRECTORY in the Glidein HTCSS configuration, was causing warnings for unknown files (PR#226)
- - Bug fix: HTCSS DC_DAEMON_LIST now is equal to DAEMON_LIST only in the Factory, in all other GlideinWMS components only selected HTCSS daemons are added explicitly to it (issue#205)
- - Bug fix: Working and local tmp directories are removed during Glidein cleanup also when the start directory is missing. This result in a loss of Glidein final status information but avoids sandbox leaks on the Worker Node. (issue#189)
- - NOTE: custom scripts should always read values via gconfig_get(). The only exception is the parsing or the line to get the add_config_line source file
- - NOTE: add_config_line is deprecated in favor of gconfig_add. Will be removed from future versions
- - NOTE: add_config_line_safe is deprecated in favor of gconfig_add_safe. gconfig_add is the recommended method to use also in concurrent scripts
- - NOTE: gconfig_add_unsafe, with a similar behavior to the former add_config_line, has been added. Its use is not recommended, unless absolutely sure that no script will be run in parallel
- - NOTE: Only the first collector in TRUST_DOMAIN is kept, following collectors are removed. This happens both in the Frontend token issuer and in the setting of the Glidein TRUST_DOMAIN (setup_x509.sh).
-
-
--
- v3_9_5 released on May 17, 2022 (Manual,
- Installation instructions)
-
- - Support for credential generator plugins
- - Support for Apptainer
- - Support for python3-rrdtool added in OSG 3.6 also for RHEL7
- - Support for CVMFS on-demand via cvmfsexec (mount/umountrepo)
- - Configurable idtokens lifetime
- - Improved frontend logging
- - Improved default SHARED_PORT configuration
- - Special handling of multiline condor config values
- - Advertise CONTAINERSW_ variables
- - Bug fix: Fixes an issue causing SciTokens transfered to the factory to be corrupted
- - Bug fix: Fixes to manual_glidein_submit for compatibility with Python 3
- - Bug fix: Fixes to OSG_autoconf for compatibility with Python 3
- - Bug fix: Fixes a minor issue in job.condor
- - Bug fix: Fixes OS detection for CentOS Stream 8
- - Bug fix: Fixes an issue with scitoken handling to gsi_proxy configured factory entries
- - Bug fix: Fixes condor tarball directory structure
- - Bug fix: Stops clearing LD_LIBRARY_PATH before starting HTCondor
- - Bug fix: Ensures that idtokens are transferred to glidein credential directory
-
-
--
- v3_9_4 released on Jan 25, 2022 (Manual,
- Installation instructions)
-
- - Includes all features and fixes of 3.7.6
- - Adds support for CentOS 8
- - Adds two new RPM packages (vofrontend-libs and vofrontend-glidein) to provide essential libraries for frontend-like applications
- - Updates license to Apache 2.0
- - Bug fix: Updates libraries for compatibility with PyYAML 6.0
- - Bug fix: Fixes a sring encoding issue affecting the cat log scripts
- - Bug fix: Fixes AWS/Google metadata encoding error
-
-
--
- v3_9_3 released on Sep 21, 2021 (Manual,
- Installation instructions)
-
- - Includes all features and fixes of 3.7.5
- - Adds type validation to the check_python3_expr.py script
- - Removes unused encodingSupport.py and its unit tests
- - Bug fix: Fixes an encoding issue affecting cloud submissions
-
-
--
- v3_9_2 released on Jun 1, 2021 (Manual,
- Installation instructions)
-
- - Includes all features and fixes of 3.7.3 and 3.7.4
- - Drops support for BDII along with its LDAP requirements
- - Adds check_python3_expr.py to validate python3 in match expressions
- - Bug fix: Fixes string encoding issues on glideFactoryLib
- - Bug fix: Fixes an issue that would cause some factory configurations to generate malformed Condor files
-
-
--
- v3_9_1 released on February 11, 2021 (Manual,
- Installation instructions)
-
- - Includes all features and fixes of 3.7.2
- - Separates httpd packages from Factory and Frontend RPMs
- - Bug fix: Frontend no longer reports duplicate glideclient daemons to the Factory
-
-
--
- v3_9 released on September 18, 2020 (Manual,
- Installation instructions)
-
- - Includes all features and fixes of 3.6.3 and 3.7.1
- - Python 3 code, some pitfalls to pay attention to are in the Bug Fix section
- - Bug fix: custom disctionaries with __getattr__ in the creation package (an infinite loop was masked in Python 2)
- - Bug fix: str vs bytes, especially in crypto libraries
- - Bug fix: None cannot be used in sorted string lists and we use (None, None, None) for unmatched jobs in the Frontend
- - Bug fix: Add __repr__ method to SubParams class
- - Bug fix: Fixed pycache of creation/lib included in multiple packages
- - Bug fix: Fixed spec file to include python3 compiled files and not python2 ones
- - Bug fix: Fixed Mapping import
- - Bug fix: Fixed str vs bytes confusion in cryptography functions and classes
-
-
--
- v3_7_7 released on June 21 2022 (Manual,
- Installation instructions)
-
- - Authenticate w/ tokens w/ the HTCondor-CE collector
- - Added CONTINUE_IF_NO_PROXY attribute to control GSI to token transition
- - Bug fix: Fixes an issue causing SciTokens transfered to the factory to be corrupted
- - Bug fix: Fixes condor_config -dump is not safe if the pool config has multiline entries
- - Bug fix: Add job.condor typo fix for bosco to 3.7.7
- - NOTE: Now branch_v3_7 is completely separate. Feature or fixes have to be backported
-
-
--
- v3_7_6 released on Jan 18 2022 (Manual,
- Installation instructions)
-
- - Flexible mount point for CVMFS in glidein
- - Frontend and factory check expiration of SciTokens
- - Shared port is default for User Collector
- - API to retrieve job information from condor jobID
- - OSG_autoconf now allows override yaml configuration files
- - expiration and other validity of JWT tokens checked before use
- - Bug Fix: condor_chirp no longer fail when python2 used in job
- - Bug Fix: IDTOKEN issuer changed from collector host to trust domain
- - Bug Fix: proxy was failing to renew when both SCITOKEN and grid_proxy auth used
- - Bug Fix: OSG_autoconf automatic assignment of submit attributes fixed
- - Bug Fix: OSG_autoconf common entry fields implemented
- - Bug Fix: OSG_autoconf now possible to append instead of replace values
- - Bug Fix: glidein clears LD_LIBRARY_PATH prior to starting HTCondor
- - Bug Fix: JOB_TOKENS no longer set in glidein environment if not needed
- - Bug Fix: IDTOKEN/SCITOKEN auth to bosco sites fixed
- - Bug Fix: improved IDTOKEN revocation procedures
- - Bug Fix: SSL added to condor *_AUTHENTICATION_METHODS
- - Bug Fix: Possible to have frontend group with only SCITOKEN authentication
- - Bug Fix: Factory stores frontend group information per-token
- - Bug Fix: Intermittent -check_v35 error fails to startup factory
- - Bug Fix: Fix glidein_starup.sh in staging area to be one with payload
- - Bug Fix: GCB support removed
- - Bug Fix: glidein_cleanup can cleanup directories chmoded read-only
- - Bug Fix: improved shellcheck testing
- - Bug Fix: corrected paths for some variables in glideins
- - Bug Fix: eliminated spurious log messages in glideins
- - Bug Fix: STARTD_JOB_ATTRS fixed in glideins
- - Bug Fix: GWMS_PATH in glidein_startup.sh corrected
- - Bug Fix: better cleanup of tmpfiles after install
-
-
--
- v3_7_5 released on Sep 2 2021 (Manual,
- Installation instructions)
-
- - Support for big files in the glideinwms repository
- - Additional download mechanism for files
- - Bug Fix: Have SciToken credential be per-entry
- - Bug Fix: Fix IDTOKEN generation in the Frontend
- - Bug Fix: Cleanup script not executing correctly at glidein termination
- - Bug Fix: Automount of CVMFS integrated into glidein script
-
-
--
- v3_7_4 released on May 10, 2021 (Manual,
- Installation instructions)
-
- - Consider not available (not Idle) partitionable slots where all GPUs are used
- - Bug Fix: Submissions from some 3.6.5 frontends to 3.7.3 factories go on hold
-
-
--
- v3_7_3 released on March 25, 2021 (Manual,
- Installation instructions)
-
- - More efficient generation of IDTOKENS by frontend
- - SciToken credentials configured and treated same as other credentials for glidein submission
- - Glideins configured without GSI but with SciToken and IDTOKENS successfully run jobs.
- - Order of places glidein checks for singularity changed
- - PATH with multiple directories can be configured to override singularity binary
- - Remove LD_PRELOAD from glidein environment by default
- - Fail without fall-back when user provided singularity image missing
- - Improved GPU implementation in glidenins
- - OSG_autoconf more configurable, auth_method and pilot_entries can be set
- - Bug Fix: Missing GSI configuration does not stop schedd query when IDTOKENS configured.
- - Bug Fix: Factory should not check condor tarball version prior to SciToken authentication to CE
- - Bug Fix: tmpfile needs to be on correct file system when factory writes tokens from frontend
- - Bug Fix: IDTOKENS auth does not fail on improper condor_version comparison
- - Bug Fix: rpm sub package (http) building issues fixed
-
-
--
- v3_7_2 released on December 21, 2020 (Manual,
- Installation instructions)
-
- - Includes all features and fixes of 3.6.6
- - Drop GLExec support
- - Mechanism for running custom scripts at the end of glidein execution
- - OSG_autoconf protection added to prevent failures when CE dissapears from OSG Collector
- - Bug fix: Frontend no longer over-generates tokens for entries
- - Bug fix: Factory does not remove glideins when idle limit is hi
-
-
--
- v3_7_1 released on November 3, 2020 (Manual,
- Installation instructions)
-
- - Includes all features and fixes of 3.6.5
- - Configuration changes to make more compatible with HTCondor 8.9 on initial install
- - SciTokens authentication between Factory and CE
- - IDTokens authentication between Factory and Frontend
- - Bug fix: Factory reconfigs and startups do not require manual creation of directories
- - Bug fix: Factory builds Condor Tarballs to send with glideins correctly
-
-
--
- v3_7 released on March 31, 2020 (Manual,
- Installation instructions,Tarball)
-
- - Includes all features and fixes of 3.6.2
- - Use of HTCondor token-auth for Glideins authentication
- - Added secure logging channel
- - Refactored glidien_startup.sh to separate out the code in heredoc sections and improve code quality
-
-
-
v3_5 released on June 7, 2019 (