Widdler is a command-line tool for executing WDL workflows on Cromwell servers. Features include:
- Workflow execution: Execute a workflow on a specified Cromwell server.
- Workflow restart: Restart a previously executed workflow.
- Workflow queries: Get the status, metadata, or logs for a specific workflow.
- Workflow result explanation: Get more detailed information on fails at the command line.
- Workflow monitoring: Monitor a specific workflow or set of user-specific workflows to completion.
- Workflow abortion: Abort a running workflow.
- JSON validation: Validate a JSON input file against the WDL file intended for use.
Widdler requires Python 2.7 and Java-1.8 to be loaded in your environment in order for full functionality to work.
Below is widdler's basic help text. Widdler expects one of three usage modes to be indicated as it's first argument: run, query, or abort.
usage: widdler.py <positional argument> [<args>]
Description: A tool for executing and monitoring WDLs to Cromwell instances.
positional arguments:
{restart,explain,log,abort,monitor,query,run,validate,label,email}
optional arguments:
-h, --help show this help message and exit
Below is widdler's run help text. It expects the user to provide a wdl file, json file, and to indicate one of the available servers for execution. The validate option validates both the WDL and the JSON file submitted and is on by default.
usage: widdler.py run <wdl file> <json file> [<args>]
Submit a WDL & JSON for execution on a Cromwell VM.
positional arguments:
wdl Path to the WDL to be executed.
json Path the json inputs file.
optional arguments:
-h, --help show this help message and exit
-v, --validate Validate WDL inputs in json file. (default: False)
-l LABEL, --label LABEL
A key:value pair to assign. May be used multiple
times. (default: None)
-m, --monitor Monitor the workflow and receive an e-mail
notification when it terminates. (default: False)
-i INTERVAL, --interval INTERVAL
If --monitor is selected, the amount of time in
seconds to elapse between status checks. (default: 30)
-o EXTRA_OPTIONS, --extra_options EXTRA_OPTIONS
Additional workflow options to pass to Cromwell.
Specify as k:v pairs. May be specified multipletimes
for multiple options. See
https://github.com/broadinstitute/cromwell#workflow-
optionsfor available options. (default: None)
-V, --verbose If selected, widdler will write the current status to
STDOUT until completion while monitoring. (default:
False)
-n, --no_notify When selected, disable widdler e-mail notification of
workflow completion. (default: False)
-d DEPENDENCIES, --dependencies DEPENDENCIES
A zip file containing one or more WDL files that the
main WDL imports. (default: None)
-D, --disable_caching
Don't used cached data. (default: False)
-S {ale1,btl-cromwell,localhost,gscid-cromwell}, --server {ale1,btl-cromwell,localhost,gscid-cromwell}
Choose a cromwell server from ['ale1', 'btl-cromwell',
'localhost', 'gscid-cromwell'] (default: None)
For example:
widdler.py run myworkflow.wdl myinput.json -S ale
This will return a workflow ID and status if successfully submitted, for example:
{'id': '2f8bb5c6-8254-4d38-b010-620913dd325e', 'status': 'Submitted'}
This will execute a workflow that uses subworkflows:
widdler.py run myworkflow.wdl myinput.json -S ale -d mydependencies.zip
Users may also invoke Widdler's monitoring capabilities when initiating a workflow. See below for an explanation of monitoring options.
If a workflow has been previously executed to a Cromwell server, it is possible to restart the workflow after it has completed and run it again with the same inputs simply by providing the workflow ID and server of the original run. The usage for performing this action is as follows:
usage: widdler.py restart <workflow id>
Restart a submitted workflow.
positional arguments:
workflow_id workflow id of workflow to restart.
optional arguments:
-h, --help show this help message and exit
-S {ale,btl-cromwell}, --server {ale,btl-cromwell}
Choose a cromwell server from ['ale', 'btl-cromwell']
(default: None)
For example:
Will restart workflow b931xxx and return the new workflow id like so:
Workflow restarted successfully; new workflow-id: 164678b8-2a52-40f3-976c-417c777c78ef
Finally, any restarted workflows will inherit the labels of it's originating workflow.
Below is widdler's query help text. Aside from the workflow ID it expects one or more optional arguments to request basic status, metadata, and/or logs.
usage: widdler.py query <workflow id> [<args>]
Query cromwell for information on the submitted workflow.
positional arguments:
workflow_id workflow id for workflow execution of interest.
(default: None)
optional arguments:
-h, --help show this help message and exit
-s, --status Print status for workflow to stdout (default: False)
-m, --metadata Print metadata for workflow to stdout (default: False)
-l, --logs Print logs for workflow to stdout (default: False)
-u USERNAME, --username USERNAME
Owner of workflows to monitor. (default: amr)
-L LABEL, --label LABEL
Query status of all workflows with specific label(s).
(default: None)
-d DAYS, --days DAYS Last n days to query. (default: 7)
-S {ale1,btl-cromwell,localhost,gscid-cromwell}, --server {ale1,btl-cromwell,localhost,gscid-cromwell}
Choose a cromwell server from ['ale1', 'btl-cromwell',
'localhost', 'gscid-cromwell'] (default: None)
-f {Running,Submitted,QueuedInCromwell,Failed,Aborted,Succeeded}, --filter {Running,Submitted,QueuedInCromwell,Failed,Aborted,Succeeded}
Filter by a workflow status from those listed above.
May be specified more than once. (default: None)
-a, --all Query for all users. (default: False)
For example:
widdler.py 2f8bb5c6-8254-4d38-b010-620913dd325e query -s -S ale
will return something like this:
[{'id': '2f8bb5c6-8254-4d38-b010-620913dd325e', 'status': 'Running'}]
and:
widdler.py query 2f8bb5c6-8254-4d38-b010-620913dd325e -m -s ale
will return a ton of information like so (truncated for viewability):
{'status': 'Running', 'submittedFiles': {'workflow': '# GATK WDL\r\n# import "hc_scatter.wdl" as sub\r\n\r\ntask VersionCheck {\r\n String gatk\r\n command {\r\n source
/broad/software/scripts/useuse\r\n use Java-1.8\r\n use Python-2.7\r\n... 'ref': '/cil/shed/sandboxes/amr/dev/gatk_pipeline/output/pfal_5/Plasmodium_falciparum_3D7.fasta'}}]}, 'submi
ssion': '2017-07-14T11:26:05.931-04:00', 'workflowName': 'gatk', 'outputs': {}, 'id': '2f8bb5c6-8254-4d38-b010-620913dd325e'}]
and:
widdler.py query 2f8bb5c6-8254-4d38-b010-620913dd325e -l -s ale
[{'id': '2f8bb5c6-8254-4d38-b010-620913dd325e', 'calls': {'gatk.MakeSampleDir': [{'shardIndex': 0, 'attempt': 1, 'stderr': '/cil/shed/apps/internal/cromwell_new/cromwell-executions/ga
tk/2f8bb5c6-8254-4d38-b010-620913dd325e/call-MakeSampleDir/shard-0/execution/stderr', 'stdout': '/cil/shed/apps/internal/cromwell_new/cromwell-executions/gatk/2f8bb5c6-8254-4d38-b010-
620913dd325e/call-MakeSampleDir/shard-0/execution/stdout'}
Below is widdler's abort usage. Simply provide the
Abort a submitted workflow.
positional arguments:
workflow_id workflow id of workflow to abort.
optional arguments:
-h, --help show this help message and exit
-S {ale,btl-cromwell}, --server {ale,btl-cromwell}
Choose a cromwell server from ['ale', 'btl-cromwell']
(default: None)
This example:
widdler.py abort 2f8bb5c6-8254-4d38-b010-620913dd325e -S ale
will return:
{'status': 'Aborted', 'id': '2f8bb5c6-8254-4d38-b010-620913dd325e'}
Running widdler.py explain will provide information at command line similar to the monitor e-mail, including workflow status, root directory, stdout and stderr information, and useful links. Usage is as follows:
usage: widdler.py explain <workflowid>
Explain the status of a workflow.
positional arguments:
workflow_id workflow id of workflow to abort.
optional arguments:
-h, --help show this help message and exit
-S {ale,btl-cromwell}, --server {ale,btl-cromwell}
Choose a cromwell server from ['ale', 'btl-cromwell']
(default: None)
This example:
python widdler.py explain b931c639-e73d-4b59-9333-be5ede4ae2cb -S ale
will return:
{'id': 'b931c639-e73d-4b59-9333-be5ede4ae2cb',
'status': 'Failed',
'workflowRoot': '/cil/shed/apps/internal/cromwell_gaag/cromwell-executions/gatk/b931c639-e73d-4b59-9333-be5ede4ae2cb'}
-------------Failed Stdout-------------
/cil/shed/apps/internal/cromwell_gaag/cromwell-executions/gatk/b931c639-e73d-4b59-9333-be5ede4ae2cb/call-ApplySnpRecalibration/execution/stdout:
[Errno 2] No such file or directory: u'/cil/shed/apps/internal/cromwell_gaag/cromwell-executions/gatk/b931c639-e73d-4b59-9333-be5ede4ae2cb/call-ApplySnpRecalibration/execution/stdout'
-------------Failed Stderr-------------
/cil/shed/apps/internal/cromwell_gaag/cromwell-executions/gatk/b931c639-e73d-4b59-9333-be5ede4ae2cb/call-ApplySnpRecalibration/execution/stderr:
[Errno 2] No such file or directory: u'/cil/shed/apps/internal/cromwell_gaag/cromwell-executions/gatk/b931c639-e73d-4b59-9333-be5ede4ae2cb/call-ApplySnpRecalibration/execution/stderr'
-------------Cromwell Links-------------
http://ale:9000/api/workflows/v1/b931c639-e73d-4b59-9333-be5ede4ae2cb/metadata
http://ale:9000/api/workflows/v1/b931c639-e73d-4b59-9333-be5ede4ae2cb/timing
Note that in this case, there were no stdout or stderr for the step that failed in the workflow.
(Requires Java-1.8, so make sure to 'use Java-1.8' before trying validation)
Widdler validation attempts to validate the inputs in the user's supplied json file against the WDL arguments in the supplied WDL file. Validation is OFF by default and so users must specify it using the -v flag if using widdler.py run. Validaton can also be performed using widdler.py validate if you wish to validate inputs without executing the workflow.
It will validate the following:
- That the value of a parameter in the json matches the same type of value the WDL expects. For example if the WDL expects an integer and the parameter supplies a float, this will be flagged as an error.
- That if the parameter is of type File, that the file exists on the file system.
- If a parameter specified in the json is not expected by the WDL.
- If a parameter contains the string 'samples_file' it's value will be interpreted as an input TSV file in which the last column of every row indicates a sample file. In this case, an existence check will be made on each sample file.
It will NOT validate the following:
- The contents of arrays. It can't tell the difference between an array of strings and an array of integers, but it can tell they are arrays, and if a parameter expects an array but is provided something else this will be logged as an error.
A note on validating WDL files with dependencies: due to the limitations of the current implementation of depedency validation, WDL file dependencies must be present in the same directory as the main WDL file and must be unzipped. Otherwise validation may not work.
Validation may also be run as a stand-alone operation using widdler.py validate. Usage is as follows:
usage: widdler.py validate <wdl_file> <json_file>
Validate (but do not run) a json for a specific WDL file.
positional arguments:
wdl Path to the WDL associated with the json file.
json Path the json inputs file to validate.
optional arguments:
-h, --help show this help message and exit
For example:
widdler.py mywdl.wdl myjson.json
If the json file has errors, a list of errors will be reported in the same way that the runtime validation reports. For example:
bad.json input file contains the following errors:
gatk.ts_filter_snp: 99 is not a valid Float.
gatk.tcir: False is not a valid Boolean. Note that JSON boolean values must not be quoted.
gatk.ploidy: 2.0 is not a valid Int.
Required parameter gatk.snp_annotation is missing from input json.
Required parameter gatk.ref_file is missing from input json.
Running 'widdler.py log' will print to screen the commands used by each task of a workflow. For example, running:
widdler.py log becb307f-4718-4d8b-836f-5780d64c4a82 -S btl-cromwell
Results in the following:
{u'hello.helloWorld': [{u'attempt': 1, u'shardIndex': -1, u'stderr': u'/btl/store/cromwell_executions/hello/becb307f-4718-4d8b-836f-5780d64c4a82/call-helloWorld/execution/stderr', u'stdout': u'/btl/store/cromwell_executions/hello/becb307f-4718-4d8b-836f-5780d64c4a82/call-helloWorld/execution/stdout'}]}
hello.helloWorld:
#!/bin/bash
tmpDir=$(mktemp -d /cil/shed/apps/internal/cromwell_new/cromwell-executions/hello/d90bf4f3-d9fb-4f07-92d9-0d46c40355f1/call-helloWorld/execution/tmp.XXXXXX)
chmod 777 $tmpDir
export _JAVA_OPTIONS=-Djava.io.tmpdir=$tmpDir
export TMPDIR=$tmpDir
(
cd /cil/shed/apps/internal/cromwell_new/cromwell-executions/hello/d90bf4f3-d9fb-4f07-92d9-0d46c40355f1/call-helloWorld/execution
echo Hello, amr
)
echo $? > /cil/shed/apps/internal/cromwell_new/cromwell-executions/hello/d90bf4f3-d9fb-4f07-92d9-0d46c40355f1/call-helloWorld/execution/rc.tmp
(
cd /cil/shed/apps/internal/cromwell_new/cromwell-executions/hello/d90bf4f3-d9fb-4f07-92d9-0d46c40355f1/call-helloWorld/execution
)
sync
mv /cil/shed/apps/internal/cromwell_new/cromwell-executions/hello/d90bf4f3-d9fb-4f07-92d9-0d46c40355f1/call-helloWorld/execution/rc.tmp /cil/shed/apps/internal/cromwell_new/cromwell-executions/hello/d90bf4f3-d9fb-4f07-92d9-0d46c40355f1/call-helloWorld/execution/rc
Widdler allows the monitoring of workflow(s). Unlike the query options, monitoring persists until a workflow reaches a terminal state (any state besides 'Running' or 'Submitted'). While monitoring, it can optionally print the status of a workflow to the screen, and when a terminal state is reached, it can optionally e-mail the user (users are assumed to be of the broadinstitute.org domain) when the workflow is finished.
Monitoring usage is as follows:
usage: widdler.py monitor <workflow_id> [<args>]
Monitor a particular workflow and notify user via e-mail upon completion. If
aworkflow ID is not provided, user-level monitoring is assumed.
positional arguments:
workflow_id workflow id for workflow to monitor. Do not specify if
user-level monitoring is desired. (default: None)
optional arguments:
-h, --help show this help message and exit
-u USERNAME, --username USERNAME
Owner of workflows to monitor. (default: <your user name>)
-i INTERVAL, --interval INTERVAL
Amount of time in seconds to elapse between status
checks. (default: 30)
-V, --verbose When selected, widdler will write the current status
to STDOUT until completion. (default: False)
-n, --no_notify When selected, disable widdler e-mail notification of
workflow completion. (default: False)
-S {ale,btl-cromwell}, --server {ale,btl-cromwell}
Choose a cromwell server from ['ale', 'btl-cromwell']
(default: None)
Aside from monitoring of a single workflow with widdler's run command, you can also execute a monitor as in the following example:
widdler.py monitor 7ff17cb3-12f1-4bf0-8754-e3a0d39178ea -S btl-cromwell
In this case, widdler will continue to silently monitor this workflow until it detects a terminal status. An e-mail will be sent to @broadinstitute.org when a terminal status is detected, which will include the metadata of the workflow.
If --verbose were selected, the user would have seen a STDOUT message indicating the workflows status at intervals defined by the --interval parameter, which has a default of 30 seconds.
If --no_notify were selected, an e-mail would not be sent.
(Note this feature is still under active development and is currently quite primitive)
User's may also monitor all workflows for a given user name by omitting the workflow_id parameter and specifying the --user parameter like so:
widdler.py monitor -u amr -n -S btl-cromwell
Here, the user 'amr' is monitoring all workflows ever executed by him using widdler. Any workflows not executed by widdler will not be monitored. Workflows in a terminal state prior to execution will have an e-mail sent immediately regarding their status, and any running workflows will result in an e-mail once they terminate. Using the --verbose option here would result in STDOUT output for each workflow that is monitored at intervals specified by --interval.
Widdler logs information in the application's logs directory in a file called widdler.log. This can be useful to find information on widdler executions including workflow id and query results and can help users locate workflow IDs if they've been lost. Each execution in the log is presented like so, with the user's username indicated in the start/stop separators for convenient identification.
2017-07-14 12:10:44,746 - widdler - INFO - Parameters chosen: {'logs': False, 'func': <function call_query at 0x00000000040B8378>, 'status': True, 'workflow_id': '7ff17cb3-12f1-4bf0-8754-e3a0d39178ea', 'server': 'btl-cromwell', 'metadata': False}
2017-07-14 12:10:44,746 - widdler.cromwell.Cromwell - INFO - URL:http://btl-cromwell:9000/api/workflows/v1
2017-07-14 12:10:44,746 - widdler.cromwell.Cromwell - INFO - Querying status for workflow 7ff17cb3-12f1-4bf0-8754-e3a0d39178ea
2017-07-14 12:10:44,747 - widdler.cromwell.Cromwell - INFO - GET REQUEST:http://btl-cromwell:9000/api/workflows/v1/7ff17cb3-12f1-4bf0-8754-e3a0d39178ea/status
2017-07-14 12:10:44,812 - widdler - INFO - Result: [{'id': '7ff17cb3-12f1-4bf0-8754-e3a0d39178ea', 'status': 'Running'}]
2017-07-14 12:10:44,813 - widdler - INFO -
-------------End Widdler Execution by amr-------------
Widdler now supports executing workflows on Cromwell servers running on Google Cloud VMs, with data being processed and stored in cloud storage buckets. To use a cloud server, the user must first authenticate manually with GCP as follows (this assumes you've already created a Google Cloud Platform account):
- Log in to Stout or an UGES interactive host.
- Type 'use Google-Cloud-SDK'.
- Type 'gcloud auth login '.
- Copy and paste the URL printed on screen into a browser and follow the steps presented.
- Paste the verification code provided into the shell window at the 'Enter Verification Code' prompt.
- You are credentials are now activated.
- Contact the GCP Bucket owner to ensure you have access to the buckets that Widdler will use to process your workflow.