RabbitMQ Cluster Operator memory resource limit 500Mi

[ilabrovic]

We have deployed the RabbitMQ-cluster-operator 2.6.0 and running into OOMKiller.
It seems to run at 500Mi memory limit hardcoded?
I can t configure the resources of this cluster operator pod since it is managed by the operator.
How can i deal with this?

[ilabrovic]

After changing defaults in the Subscription resource, we concluded we need 730Mi memory limit to get the rabbitmq-cluster-operator started. Do you have any clue why we need more than the default 500Mi limit?

[mkuratczyk]

Seems like I can reproduce this by declaring a lot of services (completely unrelated to RMQ). No such issue if there's a lot of ConfigMaps. I will look into this

[mkuratczyk]

A change that should fix this has been merged. If you have an option to give it a try, that'd be great. The image is rabbitmqoperator/cluster-operator:sha-62d64ca. Note that if you just deploy it to an empty test cluster, you won't see a difference. Assuming the problem is where we think it is, the Kubernetes cluster needs to have quite a lot of objects (earlier I thought the ConfigMap wouldn't cause that but they do - mine were just very small).

[ilabrovic]

Thanks, i will investigate if we have that option to try this image. I'll get back to you if/when we can.