ChangeLog

Summary of changes in minor versions 3.1.x:

3.1.6:
- fix build on Lustre 2.12.4
- check the filesystem returns consistent statfs values
- chglog_reader: reopen changelog in case of unexpected error
- chglog_reader: de-duplicate HSM STATE events (keep the latest)
- policy run time stat now includes initial DB request time
- fix the use of archive_id parameter in lhsm
- port to CentOS8 and lustre master (2.13.52)

3.1.5:
- Lustre: compatibility with FPL and DoM (no full-featured support)
- Fix and improvement of ratelimit scheduler
- Stop retrying SQL requests on SIGTERM
- Faster remove policies (disable sorting)

3.1.4:
* Web gui:
    - new access control criteria: IP address, hostname;
    - tasks: scheduled requests, and possiblity to keep result history;
    - custom graphs.
* Performance improvements:
    - Improved rules-to-SQL conversion engine. Enable faster policy runs.
    - Auto-tune hash table sizes (changelog reader and entry processor).
    - Pipeline tuning to reduce CPU usage and improve ingestion rate.
    - Default linking to jemalloc memory allocator for better performance.
      Note: it is advised to start MariaDB with this allocator too
      to achieve maximal performance.
    - Fix major performance bug in changelog reader (appeared in v3.1.3).
* Configuration files: support %include directives in sub-blocks.
* New parameter to 'common.copy': "mkdir=yes" creates target directories
  prior to the copy operation.
* New action 'common.move' (move an entry from one path/name to another).
* Command copytool 'lhsmtool_cmd': add man page.

3.1.3:
* Policies: add matching modes "auto_update_attrs" and "auto_update_all"
* Faster changelog reader shutdown when the process is terminated by a signal
* Changelog optimization: drop CREATE/UNLINK and MKDIR/RMDIR changelog pairs
* rbh-find:
    ° add "-links" criterion
    ° "-ost" accepts OST sets
    ° add "-iname" option for case-insensitive name matching
    ° support "-not" for links, size and dates
* FS scan optimization (use openat() to walk through the filesystem instead
  of full paths).
* Fix rules-to-SQL conversion

3.1.2:
* Implement command rbh-rebind command to assign an archived entry to a newly
  created fid (e.g. used for undelete operation).
* Make lhsm undelete more resilient to error cases
* lhsm: add tunable to allow custom/smaller UUID
* REST API:
    ° add nagios plugin
    ° add graph preview in console plugin
    ° add new filters

Main changes in robinhood 3.1:
* Ingest rate optimizations: up to x10
* New plugin type: action schedulers
  These plugins make it possible to reorder policy actions, or smooth policy execution.
  In this release, two of such plugins are provided: 'common.max_per_run' and
  'common.rate_limit'.
* Policy optimizations:
  - Improved pre-filtering: convert all policy rules to SQL for better
    pre-filtering.
  - Configurable attribute matching before/after scheduling
    (pre/post_sched_match parameters).
* FS scan: make it possible to restrict scan to several directories
  ('scan_only' configuration parameter).
  - Especially useful with Lustre's static DNE partitioning.
* Pre/post policy run commands
* Modeguard status manager
  - Enforce or clear permission bits on entries of the filesystem
* Fixes for Lustre 2.10
  - Note: this robinhood version does not support Progressive File Layouts yet
* REST API and web interface enhancements:
	- Plugin mechanism to add custom charts/fields/requests...
	- Finer access control (limit user views only to their stats)
	- All robinhood DB available through the REST API

Main changes in robinhood 3.0:
* Policies at will:
    - Define new custom policies at will, just by writting just a few lines of configuration.
    - Schedule all imaginable actions on filesystem entries.
    - Templates are provided to implement: old entries cleanup, directory cleanup, data corruption checks (by regularly checking file's checksum), Lustre/HSM policies.
* Fully configurable policies:
    - All policy aspects made configurable (scope, actions, action parameters, sort order...)
    - Fine-grained specification of policy actions and parameters.
* Plugin-based architecture:
    - Missing a specific feature in Robinhood core? Implement you own plugin to manage new kind of policies, maintain specific info in robinhood DB, enable interactions with specific site ecosystem (job scheduler, system load...), ...
* Enabling the Robinhood community:
    - Share your custom policy definitions and plugins, so that other community members can benefit from them and enrich them!
* All Robinhood features in a single instance:
    - No more distinct and incompatible robinhood "flavors" (tmpfs, lhsm, backup...)
    - All policies can be implemented in a single robinhood instance.
* Get a better overlook of your filesystem contents than ever!
    - New web interface
    - REST interface to query filesystem stats
    - Tag your filesystem entries based on arbitrary criteria using the new fileclass implementation
    - Flexible reporting with "rbh-find -printf"
* RedHat 7 integration:
    - systemd support.
    - Per filesystem service: robinhood@<fsname>
* New features for Lustre/HSM:
    - Undelete: recover accidentally deleted files.
    - Implement Lustre/HSM with any backend, using the generic command copytool shipped with robinhood: 'lhsmtool_cmd'.
    - UUID support for Lustre/HSM copytools.
* Convenient:
    - Automatic DB conversion after major upgrades.
* Even more robust & Improved code quality.

Changes between version 2.5.4 and 2.5.5:
* [lustre] Support Lustre versions up to 2.7.
* [DB performance] Allow both batching and parallelizing DB operations when
  accounting is OFF. Benchmarks show a x3~x4 speedup of DB ingest rate in this
  case.
* [DB performance] Reduce lock contention on indexes of STRIPE_ITEMS table.
* [acct] Avoid accounting stats inconsistency when this feature is enabled/
  disabled/re-enabled.
* [policy performance] Disable ignored class re-matching by default.
* [policy performance] Disable sort by mtime by default.
* [policy performance] Allow disabling sort by atime for purge policies.
* [fix] Fix segfault when log rotation occurs under heavy logging.
* [fix] When processing a RENAME changelog record, match the new path in
  filesystem instead of building it from DB (which caused 'incomplete path'
  errors).
* [fix] Fix sendfile copy operation for backup mode.
* [fix] Fix various requests about directories: --topdirs with --filter-path
  option, --toprmdir, ...
* [fix] Fix possible duplicate insert error when updated entry attributes are
  only in ANNEX_INFO table (rare).
* [fix] Some purge parameters may not be taken into account.
* [packaging] clean permissions in distribution tarball.
* [report performance] Improve performance of path matching when running
  directory reports with --filter-path option.

Changes between version 2.5.3 and 2.5.4:
* [lustre] Lustre 2.4+: detect all stripe changes and update the DB accordingly.
* [scan] Prevent from dropping entries from DB when opendir or stat fail.
* [DB] Optimization to batch more DB requests.
* [DB] Allow using any MySQL engine (new config parameter: listmanager::mysql::engine)
* [rbh-config] Avoid backup_db to lock the whole tables with innodb.
* [bugfix] Improve robustness to corrupted mtimes.
* [bugfix] Fix possible crash in db_exec_sql.
* [bugfix] Fix possible overflow when executing a custom archive command.
* [backup mode] Clean all non-printable characters in backend path.
* [pkg] libattr-devel is now mandatory to build robinhood.

2.5.4 release notes:
* "innodb = enabled/disabled": this parameter will be deprecated.
  It has to be replaced by "engine = innodb".
* (Lustre only) one of the field types changed in the DB.
  It will be automatically converted the first time rbh 2.5.4 starts.
  The conversion takes about 5s per million entries.
* Admin guides are now available online (http://robinhood.sf.net > "Online documentation")
  "odt" and "pdf" versions are no longer maintained.

Changes between version 2.5.2 and 2.5.3:
- custom purge_command (fix): fixed vulnerability to malicious file names.
- changelog processing (fix): fixed errors 'Entry has incomplete path in DB'
  in some case of rm/create patterns.
- migration policy (fix): don't trigger copy of files that no longer exist.
- rbh-config (feature): new option 'reset_acct' to rebuild accounting info.
- changelog reader (feature): new parameter 'dump_file' to dump all incoming
  changelog records to a file.
- Port to Lustre 2.6.

Changes between version 2.5.1 and 2.5.2:

- rbh-du: fixed major performance regression (since v2.5.0).
- rbh-find: fixed occasional crash.
- HSM and backup modes: fixed a risk of removing an existing entry from the backend
  (in some situations of hardlink/rename+unlink).
- backup mode: optimized sendfile()-based copy (Linux kernel >= 2.6.33).
- logs: avoid flood of log messages in case of DB connection error.
- alerts: added host name to alert mail title.
- rbh-config empty_db/repair_db: also manage/fix stored procedures.
- cosmetic: fix wrong display of purged blocks for count-based triggers.
- cosmetic: fix migration counter display.
- init script: check that 'ulimit -s' is reasonable.
- fixed build dependancies on Fedora19 and Fedora20.
- code sanity: fixed many 'coverity' warnings + a couple of minor memleaks.
- doc: details about RPM installation locations.
- doc: detail of 'backend' paramaters for backup mode.

Changes between version 2.5.0 and 2.5.1:

- entry processing (major fix): fixed deadlock when the pipeline is full
  and an entry with an unknown parent is encountered.
- purge (enhancement): start purging data from the most used OSTs.
- rbh-find (features): new options: -pool, -exec, -print, -nouser, -nogroup, -lsost
- rbh-find (optimization): automatically switch to bulk DB request mode when
  command argument is filesystem root (+new option -nobulk to disable it).
- logging (enhancement): new config parameters to control log header format
- backup (feature): allow compressing data in archive.
- backup (fix): wrong path in archive when robinhood root directory != mount point.
- backup (fix): fix segfault when importing a single file with a FID-ending name.

Changes in version 2.5.0:
Summary:
- filesystem disaster recovery features
- new namespace management (new DB schema to properly handle hardlinks, renames...)
- scanning and changelog processing optimizations
- database optimizations (requests batching)
- many other changes, improvements and code cleaning...

Details:
- rbh-diff:
    * new command to detect differences between the filesystem and the information
      in robinhood database.
    * option "--apply=fs" for disaster recovery purpose: restore the filesystem
      metadata from robinhood DB.
    * makes it possible to rebuild a Lustre MDT from scratch, or from a LVM snapshot
      (see "Robinhood Lustre disaster recovery guide" for more details).
- database:
    * new namespace implementation in database with new NAMES table (Cray contribution)
        - fixes/improves hardlink support
        - fixes/improves Lustre ChangeLog hardlink/rename/unlink support
        - saves DB storage space
    * database request batching: significantly increase database ingest rate.
      No longer needs innodb_flush_log_at_tx_commit != 1 to speed up DB operations.
    * additional information in DB that can help for disaster recovery:
      symlink info, access rights, stripe object indexes, stripe order, nlink...
    * set default commit behavior to transaction (prevent from DB inconsistencies)
    * optimized multi-table requests
    * optimization: minimized attribute set in DB update operations
      (don't update attributes that didn't change)
    * Fix: deal with mysql case insensitivity for string matching
    * triggers and stored procedures versioning mechanism
    * prevent from overflows for large INSERT requests, wide stripes...
    * prevent from DB deadlocks
- scanning:
    * --partial-scan option is deprecated and replaced by an optional argument to --scan (e.g. --scan=/fs/subdir).
    * better management of partial scans:
        - better detection of removed entries vs. entries moved from a directory to another.
        - partial scans can be used for initial DB population (even if the DB is initially empty).
    * garbage collection of removed entries in DB is a long operation when terminating a scan (and even more
      when terminating a partial scan). Added --no-gc option to skip it (recommanded for partial scans).
    * automatically enabling --no-gc if the DB is initially empty (eg. for initial scan).
    * optimization: use *at() functions (openat, fstatat) and readdir by chunk (using getdents) instead of POSIX lstat() and readdir_r().
    * optimization: use NOATIME flag to access entries as much as possible
    * optimizations of get_stripe and get_fid operations.
    * new --diff option for robinhood --scan and --readlog: output detected changes in a diff-like format.
- Lustre changelogs:
    * changelog batching (Cray contribution): to speed up changelog processing,
      robinhood retains changelog records in memory a short time,
      to aggregate similar/redundant Changelog records on the same entry before
      updating its database.
    * support multiple changelog readers (for DNE) as multiple threads (default)
      or as multiple processes, possibly on different hosts, by giving a MDT index
      to --readlog option.
    * resilience to filesystem umount/mount.
- rbh-report:
    * new option --entry-info to get all the stored information about an entry
    * option --dump-ost can now list multiple OSTs and support ranges notation (e.g. 3,5-8,12-23).
    * --dump-ost output indicates if a file has data on a given OST (could be striped on the OST but have no data on it).
- rbh-find:
    * new option -crtime to filter entries on creation time.
    * output ordering closer to find output
    * added missing info in 'rbh-find -ls' output (nlink, mode, symlink info...)
- robinhood-backup:
    * by default, use a built-in copy function to avoid the cost of forking copy commands.
    * rbh-backup-rebind: tool to rebind an entry in the backend if its fid changed in the filesystem
      for any reason (file copied to a new one to change its stripe, etc...)
    * rbh-backup-recov new features and options:
        --list (list information about entries to be recovered)
        --ost <ost_set> to only recover entries for a given set of OSTs (support range notation):
            the basic use-case is OST disaster recovery.
        --since <time> to only recover entries modified since a given date:
            the basic use case is after restoring an OST snapshot.
    * symlinks archiving to backend made optional (new parameter 'archive_symlinks')
      as they can now be restored using robinhood database information.
- configuration:
    * can specify environment variables in config file (e.g. fs_path = $ROOT_DIR ;)
    * prevent from using a wrong config file (Cray contribution):
        - only check files in /etc/robinhood.d/<purpose>, no longer in the current directory
        - fails if to many config files are available.

Changes between version 2.4.2 and 2.4.3:
* [lustre] support of Lustre 2.4
    - DNE not fully supported yet: if running multiple MDS,
      run 1 instance of changelog reader per MDT.
    - Detect file layout changes (new changelog record CL_LAYOUT).
* [lustre] added statistics about changelog processing speed.
* [policies] new parameter 'recheck_ignored_classes' to allow/avoid
    rematching entries from ignored classes in migration and purge policies.
* [web ui] security patch to prevent from SQL injection.
* [lustre] fix stack overflow when handling files with wide stripes.
* [DB] better handling of ER_QUERY_INTERRUPTED MySQL error.
* [DB] fixed DB connection leaks.
* Backup & HSM modes:
   - [fix] fix segfault in import command when uid/gid can't be resolved.
   - [rbh-report] fix bad display of total volume with -u or -g.
* Migration policy features and optimizations:
   - [feature] new parameter 'lru_sort_attr' to select LRU sort criteria for policy application.
        Previously based on last modification time, it can now be one of:
        creation, last_archive, last_mod, last_access.
   - [feature] special meaning for condition 'last_archive == 0':
        matches entries that have never been archived.
   - [feature] suspend migration if copy error rate exceed a threshold.
        This is controled by 'suspend_error_pct' and 'suspend_error_min' parameters.
   - [stats] migration stats while migration is running: added skipped and error counters.
   - [optim] avoid rechecking ignored entries at each pass
   - [optim] smoother feeding of migration workers queue
* Code & environment:
   - [build] can specify a path to alternative lustre source tree in ./configure
   - [tests] allow specifying an alternative path to lfs command

Changes between version 2.4.1 and 2.4.2:
* [general] immediate exit on ctrl+C: don't process all queued operations, just finish current.
* [general] LSB compliance if daemon is already started.
* [DB] validation with MariaDB (replacement for MySQL in Fedora19).
* [config] can set default config file using RBH_CFG_DEFAULT environment variable.
* [config] more precise message if no config file is found.
* [rbh-find] added -not/-! option to rbh-find.
* [lustre] fix for 16 chars pool names.
* [bugfix] fixed memleak in rbh-find.
* [bugfix] fixed segfault if checking scan deadline occured exactly when scan ended.
* [logs] display bandwidth and rate stats during migration run.
* [logs] fix: DB get operations were counted twice in stats.
* [cosmetic] removed "connection failed" warning for one shot commands.
* [cosmetic] fix typos in logs.
* [devel] port to automake 1.12 (since Fedora18).

Changes between version 2.4.0 and 2.4.1:
* [lustre] better file size change detection using CLOSE events
  from MDT ChangeLog (requires Lustre 2.2 or +)
* [scan] optimization: using fstatat and getdents
* [rbh-find] added -atime/-amin options
* [logs] add'l information in logs (DB operations, HSM_rm details)
* [logs] log to stderr if opening of the log file fails
* [fix] scan blocked if final DB operation failed
* [fix] avoid DB lock exhaustion for huge requests
* [backup] manage cross device rename in backend
* [backup] rebind an entry in backend after fid change (e.g. restripe)


New features in robinhood 2.4:
* rbh-du and rbh-find: "du" and "find" clones querying robinhood's database
  Faster way to search for entries in a filesystem!

  Performance comparison for a 1 million entries Lustre v2 filesystem
        find /lustre -user foo -type f -size -32M -ls
        (no possible criteria on OST index)
        > 58m13s

        lfs find /lustre -user foo -type f --obd lustre-OST0001
        (no possible criteria on size)
        > 20m46s

        rbh-find /lustre -user foo -type f -size -32M -ost 1 -ls
        > 1.2s

* Directory reporting: top directories per dirent count, per avg file size
  (useful for small file hunting)
* File size profiling: global, per user, per group, per fileclass...
  (+additional section in webUI)
* Sorting user/groups by size range (eg. percentage of files < 1G)
  (useful for small file hunting)

* Partial scans to update only a subset of the filesystem.
  => allow distributed scans by splitting the namespace into
     partial scans running on multiple clients.

Other changes in 2.4.0:
* [packaging] rpm name 'robinhood-tmp_fs_mgr' changed to 'robinhood-tmpfs'
* [packaging] 'rbh-config' command moved to new RPM 'robinhood-adm'
* [report] refurbished rbh-report output format
* [policies] new criteria on file creation time
* [database] use innodb by default for MySQL engine
* [system] ability to detect "fake mtime" (mtime != actual modification time)
* [system] improved filesystem detection,
           using fsname or devid as FS identifier (config driven)
* [scan] can trigger external completion command when a scan ends
* [misc.] can use short config name instead of full path
          (eg. "-f <name>" instead of "-f /full/path/to/name.conf")
* [backup] directory and symlink recovery
* [lustre] port to Lustre 2.2 and 2.3
* [lustre] support for new Changelog record struct (lu-1331)
* [fix] max_rm_count=0 resulted in no rm (instead of unlimited)
* [fix] segfault in realpath() on Ubuntu
* [fix] unsigned arithmetic issue with MySQL 5.5

Changes between version 2.3.3 and 2.3.4:
- Faster and safer shutdown on SIGINT/SIGTERM
- Can use short config name instead of full config file path.
  E.g. "-f myconf" instead of "-f /etc/robinhood.d/tmp_fs/myconf.cfg"
- Consider all non-dirs for classinfo (instead of files only)
- Implemented max_rm_count in hsm remove policy
- clearer messages about DB connection and retries
- added lu543 configure option (must be enabled if this patch
  is integrated to your Lustre distribution)
- Better block counting for purges
- backup/shook modes:
    - import of existing files from backend
    - entry state set to 'archive_running' during migration
    - recovery for entries with 'release pending' or 'restore running' state on startup
      (new parameter: check_purge_status_on_startup)
    - enable DB rebuild if it is lost
    - fix: symlink recovery
    - improvements of rbhext_tool_clnt/svr (timeout, traces, ...)
    - user.shook_state xattr changed to security.shook_state
      (to avoid users to change it)
- Fix: Don't consider 'released' entries for quota-like purge triggers
- Fix: migrate-group did migrate user
- Generate up-to-date template automatically at RPM installation
- 72 new regression tests (all policies conditions and config file parameters are tested)

Changes between version 2.3.2 and 2.3.3:
[webgui]
    - added FS name to page title and page header
    - added missing file in RPM (.htaccess)
[reports]
    - new options for top-users/top-groups: --by-avgsize, --count-min, --reverse
    - Lustre changelog stats in 'rbh-report -a'
[policies]
    - fix: 'tree' condition must match root entry
    - fix: migration class matching at scan time
[config]
    - simpler parameter 'scan_interval'
    - fix: don't reload config of disabled modules on SIGHUP
    - fix: on SIGHUP, don't reload parameters specified on cmd line
[database]
    - retry on connection failure
[stats]
    - dump process stats on SIGUSR1
[backup]
    - clean special chars in archive names
    - fix issues in symlink archiving
[lustre]
    - specific compilation option for jira's LU-543
[misc]
    - code cleaning, sanity checks, improved traces...


Changes between version 2.3.1 and 2.3.2:
- [webgui] Web interface (beta)
- [quota/alerts] Implemented quota alerts on inode count (users and groups)
- [reporting] New option --by-count for --top-users, to sort users by entry count
- [database] Support of InnoDB MySQL engine
- [database] MySQL 4 compatibility fix
- [bugfix](minor) handling DB deadlock error
- [bugfix](tweak) added acct parameters to default and template outputs
- [testing] big tests with 1M entries
- [backup] about backup mode (beta):
    - [bugfix](major) fixed error determining symlink status
    - [bugfix](minor) don't consider 'new' entries in deferred removal
    - [trace] display warning if mtime in FS < mtime in backend

Changes between version 2.3.0 and 2.3.1:
- [bugfix](major) Wrong accounting values if file owner changes
- [bugfix](major) SQL error for widely striped files
- [compat] Compatibility fix for MySQL servers between 5.0.0 and 5.0.32

Changes between version 2.2.3 and 2.3.0:
- [optim.] instant accounting reports (user/group usage, fs content summary, ...)
- [reporting] split user usage per group (--split-user-groups option)
- [reporting] split group usage per user (--split-user-groups option)
- [feature] new policy criteria for Lustre FileSystems: ost_index
- [reporting] detailed FS scan statistics in "rbh-report -a"
- [misc.] fast and clean abort on ctrl^c (during scan, migration and purge)
- [admin.] automatically disables features that are not defined in config file
- [admin.] "rbh-config backup_db" helper to create a robinhood DB backup
- [misc.] -V option displays Lustre version and release number
- [tweak] changed 'watermark' parameters to 'threshold'
- [tweak] changed 'notify_lw' and 'alert_hw' parameters to 'alert_low' and 'alert_high'
- [database] alternative port or socket file can be used for MySQL connection
- [database] limiting DB access rights for reporting command
- [bugfix](major) fixed inconsistent pool names
- [bugfix](minor) kill -HUP terminated the process if no trigger was defined
- [bugfix](minor) 'unknown' status not correctly filtered in '--dump-status' report
- [bugfix](tweak) added 'reload' in short help of SLES init script
- [misc.] code cleaning, error message cleaning, removed some obsolete code
- [feature] new robinhood flavor to track modifications in a Lustre v2 filesystem, and backup data to an external storage (current status: Alpha testing only).
  As part of this feature:
        - soft rm + command to retrieve removed files
        - disaster recovery command
        - "--migrate-file" option to archive a single file
        - pre-maintenance mode to smoothly backup the whole filesystem content before a due date.


Changes between version 2.2.2 and 2.2.3:

- [feature] periodic purge trigger
- [feature] options for controlling trigger notifications
- [doc] pdf documentation updated

Changes between version 2.2.1 and 2.2.2:

- [bugfix] (major) fixed "duplicate key" errors
- [bugfix] (major) FS scan sometimes blocks on Lustre 2
- [misc.] integration to automatic testing suite (Hudson)

Changes between version 2.2.0 and 2.2.1:

- [feature] new purge command: --purge-class to apply purge policy on files in a given class
- [feature] new migration command: --migrate-class to apply migration policy on files in a given class
- [feature] support of syslog for logging
- [report cmd] Added summary line to all reports, with total nbr entries and volume.
- [report cmd] Added '-q' option to hide headers and footers in reports.
- [optim.] changed primary key format to reduce DB requests
- [misc.] new command 'repair_db' in rbh-config, to fix tables after a MySQL server crash.
- [compat.] Support for Lustre MDT changelogs on Lustre v2.0 final
- [compat.] port to FreeBSD
- [admin] added 'reload' action to init.d script
- [misc.] a gap in OST index list should displays a warning, not an error
- [pkg] common spec file for both el4, el5 and el6
- [bugfix] handling large UNIX groups (>4k) and long lists of alt groups.
- [bugfix] retrieving Lustre pool fails with error "Unsupported Lustre magic number"
- [bugfix] wrong class matching on OST pools when scanning
- [bugfix] unescaped SQL strings caused error for filenames with single quotes
- [bugfix] error in init script when RBH_OPT contains several options

Changes between version 2.1.5 and 2.2.0:

- [feature] fileclass union/intersection/negation
- [feature] rbh-report displays last matched fileclass
- [feature] new reporting command '--class-info' generates fileclass summary
- [feature] new reporting option '--filter-class' to dump entries per fileclass
- [feature] alert batching: send a mail summary instead of 1 mail per matching entry
- [feature] alert improvements: named alerts, tweak changes
- [feature] special wildcard '**' in 'path ' or 'tree' conditions matches any count of directory levels
- [feature] quota-like purge triggers fully implemented (on group or user)
- [feature] triggers on used inode count in filesystem
- [feature] '--check-triggers' option to check triggers without purging files
- [feature] notification can be sent when a high watermark is reached (for triggers)
- [feature] rbh-config helper now supports batch commands
- [feature] Lustre 2.0 ready
- [optim.] configurable fileclass periodic matching to reduce calls to filesystem
- [optim.] configurable attr/path periodic update in DB to reduce calls to filesystem
- [bugfix] explicit trace when readdir fails
- [bugfix] issue when filtering on fields with NULL values in DB
- [bugfix] check migration timeout on last effective action, not on last queued entry
- [bugfix] name-based conditions complaining about missing auto-generated fields
- [bugfix] race condition when appplying policy lead to handle the same entry several times
- [bugfix] removing removed directories from database for recursive rmdir policies
- [misc.] added documented file in /etc/sysconfig for robinhood service parameters
- [misc.] changing source directory layout
- [misc.] documentation update

Changes between version 2.1.4 and 2.1.5:

- Major bug fix: incomplete database content after scan

Changes between version 2.1.3 and 2.1.4:

- New recursive rmdir policy (for TMP_FS_MGR purpose)
- changed default value for max_pending_operations
  (unlimited value could result in excessive memory usage)
- removing useless fields and redundant information in database
- rh-* commands renamed to rbh-*, to avoid conflicts and confusions
  with RedHat commands.
- check conflicting flags in configure

Changes between version 2.1.2 and 2.1.3:

- SQLite support (should only be used for testing purpose or small filesystems)
- Support of relative paths in 'path' and 'tree' conditions
- Migration timeout mechanism
- Prompting for database admin password in rh-config script

Changes between version 2.1.1 and 2.1.2:

- New reporting commands: Dump all files (--dump-all) and dump files
  by status (--dump-status).
- New configuration helper script: "rh-config"
- Made RPM relocatable
- BUG FIX: wrong scan duration when using volume-based purge triggers
- Lustre-HSM: Checking previous migrations status when restarting
- Lustre-HSM: CL_TIME record support (bz 19505)
- Lustre-HSM: multi-archive support (archive_num)
- Lustre-HSM: new --sync option (immediately archive all modified files)
- Lustre-HSM: changed --handle-events action switch to --readlog
- Fixed SLES portability issues

Changes between version 2.1.0.beta2 and 2.1.1:

- Added new report options: --dump-ost, --dump-user, --dump-group
- Added --filter-path option to reporting tool.
- TMP FS MGR purpose ported to Lustre 2.0-alpha5 (including changelog
  support).
- documentation updates (in doc/admin_guides)
- Each purpose has its own service and binary names,
  to make it possible to install and run several robinhood with
  differents purposes on the same machine.
- Added '--disable-lustre' compilation switch for disabling Lustre specific
  features
- Added '--disable-fid-support' compilation switch, to force addressing
  entries by path, not by their Lustre fid.
- Integration of new purpose "SHERPA" (software suite for cache management)
- Generated RPM name includes lustre version it was built for.
- report command displays help if is started without option.

Changes between 2.1.0.beta1 and 2.1.0.beta2:

- Extended attributes support in policy definition
- 32 bits plateforms compatibility fixes
- Fixed bug when using mysql4
- Added parameter to force changelog polling
- Fixed minor compilation warning
- commands now search for config file in /etc/robinhood.d if no config file is
  given on command line

Changes from v2.0.1 to 2.1.0.beta1:

- added '--dry-run' option instead of "simulation_mode" parameter in config file
- added '--once' option, to perform an single pass of a given policy or action
  and exit (same as '--one-shot' option).
- Compatibility fixes for MySQL 4 and 5
- Fixed dependencies on lustre include files.
- Compatibility fixes for 32 bits platforms

Lustre-HSM specific features:
- Porting to the new changelog interface (handling changelog records as
  structures instead of text, and using CHANGELOG_FLAG_FOLLOW
  and CHANGELOG_FLAG_BLOCK options)
- Adapting to changes in changelog timestamp (secs+nano instead of jiffies)
- Use fid as primary key in database schema (for better performance)
- Added calls to llapi_hsm_request() to trigger migration, release, removal
  in HSM.
- Customizable migration hints to be passed to the copytool
- Command line options to trigger manual migrations (by user, by OST...)
- Deferred removal in HSM
- Taking HSM file status into account (dirty, released, ...)
- HSM event support
- Changelog flag support (for UNLINK and HSM event)
- Added '--ignore-policies' option to perform migration/purge to all eligible
  files without checking policy conditions.

Changes between v2.0-beta2 and v2.0.1:

- New policy definition semantics, using filesets
- Multiple fileset/policy associations
- Several changes in configuration syntax, to avoid confusions
- Support of OST pool names (on Lustre) for fileset definition and policies
- Optimizations of policy application
- Added features for Lustre-HSM