Open Task Framework (OTF) is a Python based framework to make it easy to run predefined file transfers and scripts/commands on remote machines.
Currently the framework is primarily based around being able to use SSH or SFTP to communicate with remote hosts in order to manipulate files and run commands. This is done via the use of SSH keys, which must be set up in advance.
OTF has 3 main concepts for tasks. These are:
- Transfers
- Executions
- Batches
For more details, see the task types doc
OTF can be run either as an installed script, or via a docker container
Install via pip:
pip install opentaskpyThe task-run script will be added to your PATH, and you can invoke it directly.
To run via docker, use the Dockerfile to create your own base image using just the standard opentaskpy library. However if you want to install addons, you'll need to customise this, to install the additional packages first, before bundling it as a Docker image.
docker build -t opentaskpy -f Dockerfile . # Build the image
docker run --rm --volume /opt/otf/cfg:/cfg --volume /var/log/otf:/logs--volume /home/<USER>/.ssh/id_rsa:/id_rsa -e OTF_SSH_KEY=/id_rsa -e OTF_LOG_DIRECTORY=/logs task-run -t <TASK NAME> -c /cfg # Run a taskThe default opentaskpy library is only really designed to use SSH for executions and file transfers. To do this, you need to make sure that the host/container that is running the task-run script has a private RSA key, that is trusted on all remote hosts that you're running against.
An environment variable OTF_SSH_KEY can be used to define a default SSH key to use for all SSH connectivity. This can be overridden at the transfer/execution level by specifying a keyFile in the credentials section of the protocol definition.
This is an example deployment for using OTF in an AWS environment, using BMC Control-M as the job scheduler. Control-M could be replaced with any job scheduler, AWS EventBridge for example, depending on your requirements.
"Admin/Status Portal" currently does not exist, and is a placeholder for a potential further project to visualise the OTF configs, as well as including real-time job statuses from the job scheduler.
There are several ways to customise the running of tasks.
The following details the syntax of the task-run command:
usage: task-run [-h] -t TASKID [-r RUNID] [-v VERBOSITY] [-c CONFIGDIR]
options:
-h, --help show this help message and exit
--noop Do not attempt to run anything. Only load the config files to validate that they're OK
-t TASKID, --taskId TASKID
Name of the JSON config to run
-r RUNID, --runId RUNID
Unique identifier to correlate logs with. e.g. if being triggered by an external scheduler
-v VERBOSITY, --verbosity VERBOSITY
Increase verbosity
-c CONFIGDIR, --configDir CONFIGDIR
Directory containing task configurations
--noop
This is useful for validating the configuration is valid. It will not actually do anything, other than load the configuration files and validate that they are valid. N.B. This only works for executions and transfers.
-t, --taskId
This relates to the specific task that you want to run. It is the name of the configuration file to load (without the .json or .json.j2 suffix), contained under the CONFIGDIR
-r, --runId
A log aggregation mechanism to bundle logs for several steps in a batch together. e.g. if you pass in -r some_job_name, the log files for all executions and transfers will be placed under <logdir>/some_job_name/
-v, --verbosity
There are 3 levels of logging, beyond the default INFO level. These are VERBOSE1 (1), VERBOSE2 (2), and DEBUG (3).
VERBOSITY is an integer; 1, 2 or 3
-c, --configDir
The directory containing all of the config files. These are the task definition JSON files, as well as the variables Jinja2 template file.
In order for the process to run, you must have at least one task, and a variables.json.j2 file, even if it's just an empty object definition
This is a script that can be used to validate the configuration of a batch. It will check that all tasks are defined consecutively, and that there are valid dependencies. It will also check that all tasks are defined in the configDir directory.
It is run as follows:
otf-batch-validator -c /path/to/configDir -t batch-task-nameFull list of arguments are as follows:
usage: otf-batch-validator [-h] -t TASKID [-v VERBOSITY] [-c CONFIGDIR]
options:
-h, --help show this help message and exit
-t TASKID, --taskId TASKID
Name of the JSON config to run
-v VERBOSITY, --verbosity VERBOSITY
Increase verbosity: 3 - DEBUG 2 - VERBOSE2 1 - VERBOSE1
-c CONFIGDIR, --configDir CONFIGDIR
Directory containing task configurations
These are some environment variables that can be used to customise the behaviour of the application. There are some internally used variables too, but changing them without a full understanding of the code is not advised.
OTF_NO_LOG- Disable logging to file. Only log to stderr. Set to1to enableOTF_LOG_JSON- Stderr logging will be in JSON format. Set to1to enableOTF_LOG_DIRECTORY- Path under which log files are writtenOTF_RUN_ID- (meant for internal use) An aggregator for log files. When set, all log files for a run will go under this sub directory. E.g. running a batch, all execution and transfer logs will be dropped into this sub directory, rather than a directory for each task name. This is equivalent to using-ror--runIdcommand line arguments, which is generally preferred.OTF_SSH_KEY- The private SSH key to use by default for all SSH connections. This is essential when using a basic docker container to trigger OTF. If not specified, it will default to use any private SSH keys available to the user executing the application.OTF_STAGING_DIR- Staging base directory to place files before they're dropped into their final location. Default is/tmpOTF_BATCH_RESUME_LOG_DATE- Allow resuming of batch runs from a specific date. This is useful if you want to rerun a batch from a specific date, especially if the failure happens just after midnight and the date is no longer the same as the original run. Date format isYYYYMMDDOTF_VARIABLES_FILE- Override the default variables file. This is useful when you want to use the same job definitions, but point at a different environment with different for example. Multiple files can be specified comma-separated. If variables appear in more than one file, they will be resolved from the last entry found.OTF_PARAMIKO_ULTRA_DEBUG- Enables the hiddenultra_debugoption for Paramiko. This will log all SSH communications to the console, and can be very verbose, so be careful when using this. Set to1to enable (This is for SFTP only)OTF_LAZY_LOAD_VARIABLES- Enables lazy loading of variables. This will only load variables that are used by the task definition. This can be useful if you have a large number of variables, and you only need a few of them.
By default, OTF will log to a directory called logs in the current working directory. For the docker containers, unless overridden by OTF_LOG_DIRECTORY, it will write to /logs using a symlink at /app/logs
Variables can exist at a global level, and also on a per-task basis.
You must always have a variables.json.j2 file defined at the root of your configDir.
Variables can be used using the Jinja2 template syntax within task definitions. They can also be nested.
Individual tasks can have their own local variables too
By default, all variables are loaded whenever the variables file is loaded. This can be slow if you have a large number of variables, or there are variables that some tasks use that others don't. To avoid this, you can set the OTF_LAZY_LOAD_VARIABLES environment variable to 1. This will only load variables that are used by the task definition at runtime.
Sometimes you might want to override the current date, or something specific about a file transfer when you manually run a task. This can be done using environment variables.
Standard global variables can be overridden simply by setting an environment variable that matches the name of the variable you want to override e.g. export DD=01
In the log output, you'll see something like this:
Overriding global variable (DD: 05) with environment variable (01)
To override task specific values, you can use the following format in the environment variable name:
OTF_OVERRIDE_<TASK_TYPE>_<ATTRIBUTE>_<ATTRIBUTE>_<ATTRIBUTE>
e.g. OTF_OVERRIDE_TRANSFER_SOURCE_HOSTNAME
Case doesn't matter here. For attributes that are nested within an array, you can specify the array index
e.g. OTF_OVERRIDE_TRANSFER_DESTINATION_0_PROTOCOL_CREDENTIALS_USERNAME
Again this will be logged to show you that the override is being applied.
Static variables are useful, however sometimes you need to look up something a bit more dynamic, or secret, that you don't want to hard code into the variables file.
There are 2 default lookup plugins available:
- File
- HTTP JSON
The file plugin will load the content of a file into the variable e.g.
"{{ lookup('file', path='/tmp/variable_lookup.txt') }}"The HTTP JSON plugin will perform a very basic HTTP GET request, expecting a JSON response. The value to extract is defined by a jsonpath e.g.
"{{ lookup('http_json', url='https://jsonplaceholder.typicode.com/posts/1', jsonpath='$.title') }}"This will hit the typicode.com side, and extract the title attribute from from the returned JSON file
OTF will look for plugins that are either available as an installed module (under the opentaskpy.plugins.lookup namespace), or dropped in a plugins under the config directory.
An example Python module might be: opentaskpy.plugins.lookup.aws.ssm. This can then be referenced as a variable in a template like so:
"{{ lookup('aws.ssm', name='my_test_param') }}"Alternatively a lookup plugin could be placed under cfg/plugins named my_lookup.py, and used in a template:
"{{ lookup('my_lookup', name='my_param') }}"Some variables, such as access tokens or refresh tokens for OAuth might need to be updated as part of a transfer, if a new refresh_token is issued for example. This can be achieve by using the cacheableVariables attribute in the task definition.
By default only "file" based cacheable variables are supported, but like lookup plugins and addons, it is possible to add functionality for your own.
Cacheable variables should be defined using an equivalent caching plugin to the lookup plugin that is used by the template to retrieve it. For example the following task definition shows the "dummy" handler being used, where an "access_token" is retrieved using the file plugin. When the dummy handler updates the access_token, it uses the file caching plugin to write the updated variable to the same file:
{
"task_id": 1234,
"type": "transfer",
"source": {
"accessToken": "{{ lookup('file', path='/tmp/cacheable_variable.txt') }}",
"protocol": { "name": "opentaskpy.remotehandlers.dummy" }
},
"cacheableVariables": [
{
"variableName": "source.accessToken",
"cachingPlugin": "file",
"cacheArgs": {
"file": "/tmp/cacheable_variable.txt"
}
}
]
}The cacheableVariables attribute is a list of objects, each containing the following attributes:
variableName- The name of the variable to cache. This should be defined using dot notation with reference to the base of the task definition. e.g.source.accessToken. This also supports array indexing, e.g.source.destinations[0].hostname(though this is a very unlikely use case)cachingPlugin- The name of the caching plugin to use. If anything other than "file" is used, the plugin must be available in theopentaskpy.config.variablecachingnamespace, and fully qualified.cacheArgs- A dictionary of arguments to pass to the caching plugin. This should be a dictionary or arguments required by the caching plugin, excluding the updated value itself.
Below are examples of some useful variables to start with:
"YYYY": "{{ now().strftime('%Y') }}",
"MM": "{{ now().strftime('%m') }}",
"DD": "{{ now().strftime('%d') }}",
"TODAY": "{{ YYYY }}{{ MM }}{{ DD }}",
"MONTH_SHORT": "{{ now().strftime('%b') }}",
"DAY_SHORT": "{{ now().strftime('%a') }}",
"PREV_DD": "{{ (now()|delta_days(-1)).strftime('%d') }}",
"PREV_MM": "{{ (now()|delta_days(-1)).strftime('%m') }}",
"PREV_YYYY": "{{ (now()|delta_days(-1)).strftime('%Y') }}"Usage:
"fileRegex": "somefile.*{{ YYYY }}\\.txt"now is added by OTF, it will return a datetime object aligned to whatever timezone the worker is running. If you need the UTC time, then you can use utc_now instead.
Task definitions are validated using the JSON Schemas defined within src/opentaskpy/config/schemas/. These are split up to make them more readable. The top level schema for each task type is defined within the schemas.py file, one level above.
At a later date, I plan to automate the creation of the JSON schema documentation.
The task definitions themselves live under the cfg directory, and any number of sub directories under there to allow for grouping of tasks by whatever you like. Definitions can be either a plain JSON file, or a Jinja2 template. If a template is used, it must have the .j2 suffix. JSON files may also use internal variables, this is not supported when using .j2 task definitions.
Transfers consist of a source definition, and an optional destination.
The easiest way to see usage is to look at the examples under the examples directory
Below is an example of all the options available for a transfer:
{
"type": "transfer",
"source": {
"hostname": "{{ HOST_A }}",
"directory": "/tmp/testFiles/src",
"fileRegex": ".*\\.txt",
"fileWatch": {
"timeout": 15,
"directory": "/tmp/testFiles/src",
"fileRegex": "fileWatch\\.txt"
},
"logWatch": {
"timeout": 15,
"directory": "/tmp/testFiles/src",
"log": "log{{ YYYY }}Watch1.log",
"contentRegex": "someText[0-9]",
"tail": true
},
"conditionals": {
"size": {
"gt": 10,
"lt": 20
},
"age": {
"gt": 60,
"lt": 600
}
},
"postCopyAction": {
"action": "move",
"destination": "/tmp/testFiles/archive"
},
"protocol": {
"name": "ssh",
"credentials": {
"username": "{{ SSH_USERNAME }}"
}
}
},
"destination": [
{
"hostname": "{{ HOST_B }}",
"directory": "/tmp/testFiles/dest",
"permissions": {
"group": "operator"
},
"protocol": {
"name": "ssh",
"credentials": {
"username": "{{ SSH_USERNAME }}"
}
}
},
{
"hostname": "{{ HOST_B }}",
"directory": "/tmp/testFiles/dest",
"transferType": "pull",
"rename": {
"pattern": "^(.*)\\.txt$",
"sub": "\\1-2.txt"
},
"protocol": {
"name": "ssh",
"credentials": {
"username": "{{ SSH_USERNAME }}",
"transferUsername": "{{ SSH_USERNAME }}"
}
}
}
]
}An explanation of what's going on in the order it will handled:
- Tail the log file matching named:
/tmp/testFiles/src/log{{ YYYY }}Watch1.logfor lines matching containing the regexsomeText[0-9], for up to 15 seconds, before giving up. - Poll for a file matching the regex
/tmp/testFiles/src/fileWatch\.txtfor up to 15 seconds. - Find all files matching the regex
/tmp/testFiles/src/.*\.txt, with the conditions that the are >10B and <20B in size, as well as being older than 60 seconds, but newer than 600 seconds since last modification - Transfer the files that were found to 2 destinations.
- The first destination is a simple SCP from
HOST_AtoHOST_Bwhere the file is placed under/tmp/testFiles/dest. The group ownership of the file(s) is then set tooperator - The second destination is done via a pull from the destination server into the same directory. The SCP connects to
HOST_AastransferUsername. Once the file has been retrieved, it is renamed using the following regex match^(.*)\.txt$and substitution\1-2.txt
- The first destination is a simple SCP from
- Transferred files are moved into
/tmp/testFiles/archiveonHOST_A
Executions are the simplest task to configure. They consist of a list of hosts to run on, the username to run/connect as, and the command to run.
Executions do not currently have a timeout, so can in theory run forever, or until they are killed. If a timeout is required, either use a wrapper script on the host the command is being executed on, or they should be wrapped inside a batch.
{
"type": "execution",
"hosts": ["{{ HOST_A }}"],
"directory": "/tmp/testFiles/src",
"command": "touch touchedFile.txt",
"protocol": {
"name": "ssh",
"credentials": {
"username": "{{ SSH_USERNAME }}"
}
}
}The above is running the command touch touchedFile.txt on {{ HOST_A }}, from within the /tmp/testFiles/src directory.
If multiple hosts are defined, a thread is spawned in parallel for each host. If the command fails on any of the hosts in the list, it will cause the task run to fail, once all processes have returned a result.
Batches are a little more complex. They do not contain any task definitions, only the list, and order of execution for each task.
A batch task can have multiple options set that determine the execution order and conditions, as well as how failures and task reruns are handled.
Each task in a batch has an order_id, this is a unique ID for each task, and is primarily used to define dependencies.
dependencies can be applied to any task, and are simply a list of other tasks that must be completed before it is triggered.
continue_on_fail is a boolean that defines whether a failure would cause the whole batch to fail immediately (after existing tasks have finished executing), or whether following steps get run. This defaults to false if not defined.
retry_on_rerun is a boolean that determines whether a successful task is run a second time following a failed run. If a batch exits with a failure, and then the script is reun later on that same day, by default only the steps that failed will be run. All steps can be forced to run by setting this to true
timeout specifies the number of seconds a task is allowed to be running before it gets terminated. This counts as a failure. The default timeout, if not specified is 300 seconds.
As a batch task runs, it writes out the status of each sub task to it's log file. If a failure occurs, and the batch is rerun with the same arguments, it will attempt to resume from the point of failure. To determine the previous state, the batch handler will look at only logs that are from the current date. This is tp ensure that if something failed at 1am yesterday, but hasn't been rerun, we won't try to recover from the point of failure. Sometimes you might want to recover regardless, this can be done by passing in the date of the log files that you want to recover from, using the environment variable OTF_BATCH_RESUME_LOG_DATE in the format YYYYMMDD. This will instruct the batch handler to look at logs with that date instead.
This repo has been primarily configured to work with GitHub Codespaces devcontainers, though it can obviously be used directly on your machine too.
Dev and runtime packages are defined via pipenv, with a requirements.txt for the runtime package requirements
- Clone this repo
- pip install pipenv
- pipenv --python 3.10 && pipenv install && cd src && pipenv install --editable .
python3 -m build
python3 -m twine upload --repository testpypi dist/*Here's a list of official addons:
Provides transfer and execution addons:
- Remote handler for interacting with AWS S3 buckets
- Remote handler for executing AWS Lambda functions
Lookup plugins:
- Support for AWS SSM Parameter Store for retrieving global variables
Caching Plugins:
- Support for caching variables to AWS Secrets Manager/Parameter Store
Lookup plugins:
- Support for lookup of global variables from Hashicorp Vault
Provides transfer addons:
- Remote handler for interacting with Microsoft SharePoint Online
When using external libraries, be sure to add opentaskpy.otflogging.set_log_file("<PACKAGE>") somewhere to ensure that any additional logs from the 3rd party library make their way into the OTF logs.
Plugins are very simple. They simply need a run function, and to return the required variable based on a list of kwargs provided within the config template.
3 examples are bundled by default. The simplest of which is the random plugin, which takes 2 numbers and returns a random number between the 2.
They should either be installed as a python package and accessible to load as a module, or stored within the config directory under a plugins directory.
They must sit under the opentaskpy.plugins.lookup namespace. The filename must match the name of the plugin being references from the template lookup function.
Addons allow you to write your own interfaces with other types of remote systems. This could be a custom database provider to allow you to run stored procedures on demand, or something like AWS to perform custom transfers
Addons can either be a transfer type, or execution, and must follow the same rules. They should implement all of the functions in the abstract RemoteTransferHandler or RemoteExecutionHandler class, or return a NotImplementedError exception.
You should also ensure that you define an appropriate JSON schema, and include those under the opentaskpy.addons.XXXX.remotehandlers.schemas directory
Addons can technically be named under any namespace, but must be referred to using the full package name in the task definition e.g. opentaskpy.addons.aws.remotehandlers.lambda.LambdaExecution