[YUNIKORN-1530] remove log spew in schedulerplugin

• Type: Improvement
• Status: Closed
• Priority: Major
• Resolution: Fixed
• Fix Version/s: 1.3.0
• Component/s: shim - kubernetes
• Labels: pull-request-available
• Target Version: 1.3.0
• Assignee: Krishna Birla
• Reporter: Wilfred Spiegelenburg
• Created: 06/Jan/23 05:29
• Updated: 20/Jun/23 14:41
• Resolved: 12/Feb/23 23:32

Description

In the yunikorn-k8shim/pkg/schedulerplugin/scheduler_plugin.go we log at an Info level each time we get an application that does not contain an applicationID.

This causes a log spew in large clusters as the plugins can be called for each node logging lots of lines like:

2022-12-01T06:51:19.922Z    INFO    schedulerplugin/scheduler_plugin.go:140    Skipping pod namespace1/job-27831291-mkv55 in the filter plugin because no applicationID is defined

This message should either be only logged once, using a local cache to track that the message has been logged already or moved to a Debug level. This applies to both Skipping pod messages in the code.

We should also fix up the way the messages are created and not use the fmt.Sprintf() call but leverage the proper zap logger as we do for all other messages.

Issue Links

• links to: GitHub Pull Request #528