Operator showing OOM warnings/errors

[saloni-2404]

How can we better configure the operator to avoid the following issue?

2020-12-03T16:47:27.089+0000: 0.104: [GC (Allocation Failure) 2020-12-03T16:47:27.089+0000: 0.104: [DefNew: 2176K->255K(2432K), 0.0015106 secs] 2176K->448K(7936K), 0.0015588 secs] [Times: user=0.00 sys=0.00, real=0.00 secs] and occasionally it dies via OOM

Strimzi operator runs out of memory with the default config, looks like ~100 restarts per day

NAME READY STATUS RESTARTS AGE

strimzi-operator-7fc59d96f-nssp5 1/1 Running 1110 10d
strimzi-operator-7fc59d96f-nssp5 0/1 OOMKilled 1110 10d
strimzi-operator-7fc59d96f-nssp5 0/1 CrashLoopBackOff 1110 10d
strimzi-operator-7fc59d96f-nssp5 0/1 Running 1111 10d

[scholzj]

How many resources do you have and what operator version are you using?

In general, you can try to give it more resources by editing the cluster-operator deployment. You can also use the environment variable JAVA_OPTS to pass it some specific Java options such as -Xmx etc.

[saloni-2404]

Operator version is: 2.2.20
Resources- it was having this problem on a small cluster : 3 brokers 3 zk 1 MM 1 MM2

[scholzj]

There is no 2.2.20 ... id you mean 0.20.0 I guess? That is weird. But as I suggested above, you can try to tune the resources.

[saloni-2404]

Sorry, I meant 0.17.0 version

[scholzj]

Ok. We for sure did some improvements to the memory handling in 0.20.0. So it might be actually better in the latest version.

[saloni-2404]

There are two use cases-
a) smaller cluster 4 brokers 3 zk, monitoring and not much else.
b) large cluster 12 brokers , 3zk 1x mm 3x mm2 and monitoring
How much resources should we be ideally giving in deployment-operator.yaml?
Current one is:
resources:
limits:
cpu: 1000m
memory: 256Mi
requests:
cpu: 200m
memory: 256Mi

[scholzj]

I think we configure the operator now with 384Mi or 512Mi

[saloni-2404]

ok thanks Jakub