Purpose
Pegasus WMS is primarily a NSF funded project as part of the NSF SI2 track. The SI2 program focuses on robust, reliable, usable and sustainable software infrastructure that is critical to the CIF21 vision. As part of the requirements of being funded under this program, Pegasus WMS is required to gather usage statistics of Pegasus WMS and report it back to NSF in annual reports. The metrics will also enable us to improve our software as they will include errors encountered during the use of our software.
Associated Condor Ticket that covers the development of this feature in DAGMan
https://htcondor-wiki.cs.wisc.edu/index.cgi/tktview?tn=3532,4
DAGMan Metrics
All the metrics are sent in JSON format to a server at USC/ISI over HTTP.
http://metrics.pegasus.isi.edu/metrics
If condor team, wants a maintain a separate server or URL that is an option
Metrics can be turned off, on basis of an environment variable.
The proposal is to send metrics by Condor DAGMan whenever it exits
Proposed Metrics to be reported by DAGMan
Below are common metrics, that are shared with what pegasus-plan also reports currently
| JSON KEY | DESCRIPTION |
|---|---|
| client | the name of the client ( e.g "pegasus-plan") |
| version | the version of the client |
| type | type of data - "metrics" |
| start_time | start time of the client ( in epoch seconds with millisecond precision ) |
| end_time | end time of the client ( in epoch seconds with millisecond precision) |
| duration | the duration of the condor_dagman |
| exitcode | the exitcode with which the dagman exits for a workflow |
| wf_uuid | the uuid of the executable workflow. It is generated by pegasus-plan at planning time. Can be null |
root_wf_uuid | the uuid of the root workflow in case of hierarchal workflows. It is generated by pegasus-plan at planning time. Can be null |
In addition, DAGMan we propose DAGMan send the following metrics
| JSON KEY | DESCRIPTION |
|---|---|
| jobs | the number of vanilla jobs in the input DAG file |
| dag_jobs | the number of DAG jobs in the input DAG file i.e point to another DAG executed by another instance of DAGMan |
| total_jobs | the total number of jobs in the input DAG file |
| jobs_succeeded | the number of succeeded jobs/nodes in the workflow. don't count the DAG Nodes. ( include retries?) |
| jobs_failed | the number of failed jobs/nodes in the workflow. don't count the DAG Nodes. ( include retries?) |
| dag_jobs_succeeded | the number of DAG jobs that succeeded ( include retries?) |
| dag_jobs_failed | the number of DAG jobs that failed. ( include retries?) |
| total_jobs_run | the total number of jobs runs executed in a DAG. Should be equal to jobs_succeeded + jobs_failed + dag_jobs_succeeded + dag_jobs_failed |
1 Comment
Karan Vahi
Meeting notes with Kent
To enable DAGMan metrics reporting