UPSTREAM: 40417: Always detach volumes in operator executor #13251
Conversation
Instead of marking a volume as detached immediately in Kubelet's reconciler, delegate the marking asynchronously to the operator executor. This is necessary to prevent race conditions with other operations mutating the same volume state. An example of one such problem: pod is created, volume is added to desired state of the world reconciler process starts reconciler starts MountVolume, which is kicked off asynchronously via operation_executor.go MountVolume mounts the volume, but hasn't yet marked it as mounted pod is deleted, volume is removed from desired state of the world reconciler reaches detach volume section, detects volume is no longer in desired state of world, removes it from volumes in use MountVolume tries to mark mount, throws an error because volume is no longer in actual state of world list. After this, kubelet isn't aware of the mount so doesn't try to unmount again. controller-manager tries to detach the volume, this fails because it is still mounted to the OS. EBS gets stuck indefinitely in busy state trying to detach.
gnufied
requested review from
ajsmith and
jsafrane
and removed request for
ajsmith
|
[test] |
|
LGTM for the code, however... shouldn't there be a link to BZ? |
|
@childsb or @eparis please sign off on this from a risk perspective @gnufied I see where this was targeted for 1.5 upstream picks but didn't see the actual 1.5 pick. Since the PR was from an external contributor we may want to follow up on that. |
|
@pweil- okay I will follow up on original PR for cherry pick. I think for now - lets hold off on merging this. I will drop the milestone target. |
|
I am going to push for this to get merged. Upstream PR for cherry picking to 1.5 was opened - kubernetes/kubernetes#41555 |
|
[test] but like jan said, where's the bZ? |
|
Evaluated for origin test up to e34c005 |
|
continuous-integration/openshift-jenkins/test FAILURE (https://ci.openshift.redhat.com/jenkins/job/test_pull_request_origin/239/) (Base Commit: 4bdf7ce) |
|
[merge] test flaked on #9309 |
|
Evaluated for origin merge up to e34c005 |
|
continuous-integration/openshift-jenkins/merge SUCCESS (https://ci.openshift.redhat.com/jenkins/job/merge_pull_request_origin/101/) (Base Commit: 0eb20f4) (Image: devenv-rhel7_6072) |
|
The openshift bug https://bugzilla.redhat.com/show_bug.cgi?id=1429318 Also upstream bug - kubernetes/kubernetes#32881 Having said that, the bug reported in bugzilla doesn't have reproducible steps. This is bit of a problem because - we don't know why BZ 1429318 happened. We do know, we have an existing race condition in volume detach/umount code path so we are fixing that here. But we don't know for sure, if that will fix BZ 1429318. I will keep BZ open for now, until we observe if it is really fixed. |
|
I need a BZ, in MODIFIED, which explains what we changed and why. For QA to try to pound on what changes you made at the last minute. |
|
@eparis updated the issue description with appropriate BZ. |
This fixes a important race condition in mounting a volume and if pods gets deleted while mount was in progress.
BZ- https://bugzilla.redhat.com/show_bug.cgi?id=1432949