Clster restart failed,how many resource(cpu/memory) should I give broker

I have used zeebe for a fow month.

We fact a problem recently. Can someone help me ? wo have cast a few days to solve it .

We deploy zeebe cluster in eks(aws kubenetes service) ,use zeebe helm 。And this is my cluster config:

clusterSize: “3”
partitionCount: “3”
replicationFactor: “3”
cpuThreadCount: “2”
ioThreadCount: “2”

this is my resources config for each broker:

cpu: 1
memory: 1.5Gi
cpu: 2000m
memory: 3Gi

our node in k8s only has 2core , 8G. All this three broker is scheduled in one node.

workflow instance in our system is no more than 200.I think our cluster is running in a low workload.

but when i use helm to upgrade clster , broker can’t become ready .

So, did cpu & mem that I give to broker is to small ,so when cluster restart ,they can,t become ready.

in the end ,forgive my poor english.

Hi @PigeonCai, yes - broker recovery can cause OOM as memory balloons. See:

In the version you are running, restarting pods requires more memory than running the pods.

Future releases will include the latest fixes for this.


tks for your reply!

maybe i should allocate more memory to broker .

1 Like

@PigeonCai stay tuned a new release is coming and I will update the charts this week.

1 Like