Context
There is apparently a memory leak in ocm-controller 0.27.3
NOTE: This hasn't been released yet as the current ocm-controller code needs a new release of OCM and there is only an RC available.
Version
v0.27.3 ( not released yet )
To Reproduce
Steps to reproduce the behavior:
Not sure. In our clusters we currently have 100 objects that the ocm-controller must reconcile. That is a combination of cv, localizations, cfg, resources and fluxdeployers.
With 512M requested and a 1G limit the ocm-controller can run for about 10m.
With 1G requested and a 2G limit it ran for 97m before being oom killed.
While the logs show a few errors these aren't themselves indicators or a memory problem. Instead they are about a couple of mutation objects that refer to a cv that don't exist.
Actual behavior
With the original memory settings the ocm-controller pretty quickly makes it way up to the 1G limit. I will attach a graph of this.
Expected behavior
Don't get oom killed
Screenshots (optional)
Environment
Additional Comments
Any observation, gut feeling or other information you think help the community to troubleshoot the problem?
Context
There is apparently a memory leak in ocm-controller 0.27.3
NOTE: This hasn't been released yet as the current ocm-controller code needs a new release of OCM and there is only an RC available.
Version
v0.27.3 ( not released yet )
To Reproduce
Steps to reproduce the behavior:
Not sure. In our clusters we currently have 100 objects that the ocm-controller must reconcile. That is a combination of cv, localizations, cfg, resources and fluxdeployers.
With 512M requested and a 1G limit the ocm-controller can run for about 10m.
With 1G requested and a 2G limit it ran for 97m before being oom killed.
While the logs show a few errors these aren't themselves indicators or a memory problem. Instead they are about a couple of mutation objects that refer to a cv that don't exist.
Actual behavior
With the original memory settings the ocm-controller pretty quickly makes it way up to the 1G limit. I will attach a graph of this.
Expected behavior
Don't get oom killed
Screenshots (optional)
Environment
Additional Comments
Any observation, gut feeling or other information you think help the community to troubleshoot the problem?