Re: [HTCondor-users] DiskUsage

[Jaime Frey <jfrey@xxxxxxxxxxx>]

On Jun 8, 2023, at 3:51 PM, Vibhute, Ajay M. <avibhute@xxxxxxxxxxx> wrote:

We are trying to monitor the disk utilisation by a condor job on execute node. We are using "condor_q -af DiskUsage" (on AP) to get the disk utilisation at that moment. 

However, we have noticed that the DiskUsage is updated synchronously only if "should_transfer_files   = Yes" set in the submit file. Does anyone have an idea about how often Schedd contacts Startd to get the updates? 

HTCondor monitors the jobâs disk usage only if file transfer was used (and thus the job is running under HTCondorâs managed EXECUTE directory). File transfer occurs when should_transfer_files is set to YES in the submit file. It can also happen if should_transfer_files is set to IF_NEEDED (the default) and HTCondor determines that the submit and execute machines donât have a common shared filesystem.

Stats about the jobâs execution (disk usage, memory usage, etc) are sent from the starter to the shadow every 5 minutes (STARTER_UPDATE_INTERVAL can be used to change this). The shadow sends those updates to the schedd every 15 minutes (SHADOW_QUEUE_UPDATE_INTERVAL can be used to change this).

 - Jaime