https://rancher.com/ logo
Title
f

flaky-coat-75909

08/01/2022, 11:32 PM
If I understand this line https://github.com/longhorn/longhorn-manager/blob/b88ebf82936c7335c6a3855ab44a39a2bb790d8b/controller/controller_manager.go#L169 I should setup
EngineManagerCPURequest int `json:"engineManagerCPURequest"
to positive value just add to my
kind: Node
Node.spec.engineManagerCPURequest: 200 And a node is come from
<http://longhorn.io/v1beta2|longhorn.io/v1beta2>
nodes
Am right?
but how can I limit it ?
f

famous-shampoo-18483

08/02/2022, 12:56 AM
Currently, there is no setting limiting it since it’s not recommended. Instance manager lacking of CPU may lead to IO error. The setting you mentioned is to guarantee CPU resource for instance managers.
f

flaky-coat-75909

08/02/2022, 12:58 PM
@famous-shampoo-18483 hmm it is quite understandable but sometimes I'm notice the instance-manager is using 100% cpu without reason I have posted the issue in slack https://rancher-users.slack.com/archives/CC2UQM49Y/p1658223767108449 but not receiving any response this is reason why I want to limit the instance-manager, but it looks like it is not supported because operator have diffrent logic the worst thing for me is I do not know how can I debug it when it happen it look like random issue for me
f

famous-shampoo-18483

08/08/2022, 4:17 AM
I just checked the logs you provided. Unfortunately it didn’t contain why the engine or replica got failed… Typically, the instance manager high CPU is caused by the heavy IO (to the engines or the replicas on this node), hence the disk IO saturation is high as well.