Skip to content

memory leak in ocm controller 0.27.3 #778

@dee0sap

Description

@dee0sap

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)

Image

Environment

  • windows
  • linux
  • [] mac

Additional Comments
Any observation, gut feeling or other information you think help the community to troubleshoot the problem?

Metadata

Metadata

Assignees

Labels

area/ipceiImportant Project of Common European Interestkind/tasksmall task, normally part of feature or epic

Type

Projects

Status

⌛ Wait for QA / External Feedback

Milestone

No milestone

Relationships

None yet

Development

No branches or pull requests

Issue actions