KBEA-00134 - Retry failed jobs on the agent side

Article ID:360033188471
1 minute readKnowledge base

Summary

Sometimes an ElectricAccelerator build has some intermittent failure, e.g. unable to access a remote shared file due to network connection issues. The user wants the failed jobs to be retried on the agent side to reduce the possibility of causing the whole build to fail.

Solution

You can set up an agent-side breakpoint and trigger it under specific conditions. Breakpoint is a feature you can use to debug thorny build problems.

Examples

  1. Deploy the attached retryJob.php file to your Cluster Manager in the same directory as builds.php. This is to make sure that it retries only 3 times instead of falling into an infinite loop. Please read and make necessary changes, i.e. the paths to your lock file on your Cluster Manager machine.

  2. Make any necessary changes to the attached setupBreakPoint.sh script and run it on Unix/Linux or a Cygwin environment that has access to the Cluster Manager and cmtool installed.

  3. Run the your emake command with the --emake-poison-pill-count=4 option to make sure the retry on the agent side is not terminated too early by the poison pill killer.

Applies to

  • Product versions: All

  • OS versions: All