You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
I think it's not necessary to give this pod a QoS class of Guaranteed to avoid being killed by OOM.
When the memory pressure of a node is high, it's meaningless to discover new or recycle old local volumes on that node. We can delay the operations of local volumes until the node recovers.
Ah, I forgot one thing that it's not recommended to give DaemonSet pods a QoS of BestEffort because DaemonSet controller will recreate pods immediately. When a node is under pressure, DaemonSet pods may be evicted and rescheduled back repeatedly.
We should give this Pod a QoS class of Guaranteed to avoid being killed by oom firstly.
The text was updated successfully, but these errors were encountered: