Reconcile Sync period modifications & ListVolume enhancements to support 120k CNS Volumes#3909
Conversation
|
[APPROVALNOTIFIER] This PR is APPROVED This pull-request has been approved by: chethanv28 The full list of commands accepted by this bot can be found here. The pull request process is described here DetailsNeeds approval from an approver in each of these files:
Approvers can indicate their approval by writing |
25247c3 to
e204a90
Compare
|
Triggering CSI-WCP Pre-checkin Pipeline for this PR... Job takes approximately an hour to complete |
|
SUCCESS --- Jenkins Build #1062 |
|
SUCCESS --- Jenkins Build #1064 |
|
Triggering CSI-WCP Pre-checkin Pipeline for this PR... Job takes approximately an hour to complete |
|
Triggering CSI-WCP Pre-checkin Pipeline for this PR... Job takes approximately an hour to complete |
|
Triggering CSI-WCP Pre-checkin Pipeline for this PR... Job takes approximately an hour to complete |
|
Triggering CSI-WCP Pre-checkin Pipeline for this PR... Job takes approximately an hour to complete |
|
Triggering CSI-WCP Pre-checkin Pipeline for this PR... Job takes approximately an hour to complete |
|
Triggering CSI-WCP Pre-checkin Pipeline for this PR... Job takes approximately an hour to complete |
|
Triggering CSI-WCP Pre-checkin Pipeline for this PR... Job takes approximately an hour to complete |
74e5c54 to
709fb36
Compare
|
Triggering CSI-WCP Pre-checkin Pipeline for this PR... Job takes approximately an hour to complete |
|
FAILED --- Jenkins Build #1168 |
… API in the wcp controller to look for volumeattachment status for Pod VM PVC & PV
709fb36 to
7776c51
Compare
What this PR does / why we need it:
The csi-attacher sidecar's VA reconciler runs on a default interval of 1 minute, calling ListVolumes on the CSI driver at every cycle. This flooded the CSI controller logs with noise at a rate of once per minute per cluster.
Additionally, the Supervisor (wcp) ListVolumes implementation accounted for Pod VM workloads, VM Service & VKS VM Service. Pod VMs are ephemeral VMs running inside supervisor namespaces whose disks do not appear in the standard ESXi host → VM hardware scan used to derive published nodes. As a result, PVCs consumed by Pod VMs were silently omitted from ListVolumes responses, giving the csi-attacher an incomplete view of volume-to-node mappings during its reconciliation cycles.
Testing done:
Running e2e pipelines. Results TBA
ListVolume invocation frequency is now reduced.
Before (invoked every minute):
After (invoked every 5 minutes):
Special notes for your reviewer:
Release note: