Added new retention policy "onJobFailure" #1265
Conversation
This new retention policy prevents the deletion of a pod in case of a failed build result by the job that was using the pod. This is useful for debugging failed jobs because you are still able to open a terminal in any pod and investigate log files or similar, without having to set the retention policy to 'Always' or using workarounds like long timeouts.
|
Any update on this? |
| package org.csanchez.jenkins.plugins.kubernetes.pod.retention; | ||
|
|
||
| import hudson.Extension; | ||
| import hudson.model.*; |
There was a problem hiding this comment.
Please set your IDE to avoid wildcard imports.
| // small convenience function | ||
| private void LOG(Level level, String message) { | ||
| LOGGER.log(level, () -> MODULENAME + ": " + message); | ||
| } |
There was a problem hiding this comment.
Unneeded, the class name is already printed as part of the standard formatter.
| @Override | ||
| public boolean shouldDeletePod(KubernetesCloud cloud, Pod pod) { | ||
| if (cloud == null || pod == null) { | ||
| LOG(Level.INFO, "shouldDeletePod called without actual cloud and pod"); |
There was a problem hiding this comment.
Demote all this logging to FINE.
| Jenkins jenkins = Jenkins.getInstanceOrNull(); | ||
| if (jenkins == null) { | ||
| LOG(Level.INFO, "Couldn't get the current Jenkins reference"); | ||
| return true; | ||
| } |
There was a problem hiding this comment.
| Jenkins jenkins = Jenkins.getInstanceOrNull(); | |
| if (jenkins == null) { | |
| LOG(Level.INFO, "Couldn't get the current Jenkins reference"); | |
| return true; | |
| } | |
| Jenkins jenkins = Jenkins.get(); |
| if (parts.length < 3) { | ||
| LOG(Level.INFO, "runUrl has unknown format: " + runUrl); | ||
| return null; | ||
| } |
There was a problem hiding this comment.
This is way too simplistic. If you need to look up the run, I would advise instead to add new annotations (one for the item full name, another for the build id)
|
Conflict with #1304 |
| @@ -6,6 +6,7 @@ | |||
| <ol> | |||
| <li>Never - always delete the agent pod.</li> | |||
| <li>On Failure - keep the agent pod if it fails during the build.</li> | |||
| <li>On Job Failure - keep the agent pod if the build itself fails.</li> | |||
There was a problem hiding this comment.
It should rather be called On Build Failure.
There was a problem hiding this comment.
Also, in Jenkins terminology fails means FAILURE—the build did not run to normal completion. UNSTABLE (ran to completion but with test failures) would be considered “successful”. If you mean to use the current impl (checking for SUCCESS) then be clear that an unstable build will also retain the pod.
This new retention policy
onJobFailureprevents the deletion of a pod in case of a failed build result by the job that was using the pod. This is useful for debugging failed jobs because you are still able to open a terminal in any pod and investigate log files or similar, without having to set the retention policy to 'Always' or using workarounds like long timeouts.The logic is comparatively simple: Use the
runUrlpod annotation to query the particular run from Jenkins and check its result. Sometimes the result isn't available yet, in which case we wait up to 30 seconds.To prevent having to import Jenkins classes into the test code, I'm only testing the utility methods.
This fixes JENKINS-688969