-
Notifications
You must be signed in to change notification settings - Fork 42.9k
Hidden performance problem inside volumeManager #63274
Copy link
Copy link
Closed
Labels
kind/bugCategorizes issue or PR as related to a bug.Categorizes issue or PR as related to a bug.kind/featureCategorizes issue or PR as related to a new feature.Categorizes issue or PR as related to a new feature.lifecycle/rottenDenotes an issue or PR that has aged beyond stale and will be auto-closed.Denotes an issue or PR that has aged beyond stale and will be auto-closed.sig/storageCategorizes an issue or PR as relevant to SIG Storage.Categorizes an issue or PR as relevant to SIG Storage.
Description
Metadata
Metadata
Assignees
Labels
kind/bugCategorizes issue or PR as related to a bug.Categorizes issue or PR as related to a bug.kind/featureCategorizes issue or PR as related to a new feature.Categorizes issue or PR as related to a new feature.lifecycle/rottenDenotes an issue or PR that has aged beyond stale and will be auto-closed.Denotes an issue or PR that has aged beyond stale and will be auto-closed.sig/storageCategorizes an issue or PR as relevant to SIG Storage.Categorizes an issue or PR as relevant to SIG Storage.
Is this a BUG REPORT or FEATURE REQUEST?:
/kind bug
What happened:
Currently
Kubeletsynchronizes each pod periodically, inside each loopVolumeManagerreprocesses each pod's volumes. If a volume refers to a PVC, it will fetch the latest PV/PVC object from apiserver, see here and here. So it will increase the load on apiserver/etcd if cluster scale becomes large.A possible solution is introducing a PV/PVC cache with TTL, like what ConfigMap and Secret do.