Loading...

XML

Word

Printable

Type: Bug
Resolution: Done
Priority: Major
Fix Version/s: 2.12.0.GA
Affects Version/s: 2.11.0.GA
Component/s: hosted che: telemetry, kubernetes-image-puller, devsandbox
Labels:
None

Blocked:
False
Ready:
False
Steps to Reproduce:
Hide

Create 10+ non-ephemeral workspaces for a user

Start one of the workspaces so that a common PVC is created in the user's namespace

Stop the workspace from step 2

Delete all workspaces simultaneously (please see screenshot)

In the user's namespace, there will be "rm" pods stuck in the "Pending" state.
Show
Create 10+ non-ephemeral workspaces for a user Start one of the workspaces so that a common PVC is created in the user's namespace Stop the workspace from step 2 Delete all workspaces simultaneously (please see screenshot) In the user's namespace, there will be "rm" pods stuck in the "Pending" state.
Market:

SFDC Cases Counter:
SFDC Cases Links:

Bug:
Related upstream issue [1].
When deleting all workspaces simultaneously, "rm" pods are stuck in the "Pending" state due to terminating PVC, therefore the pods are failing to start.

Expected to happen:
There should be no rm pods stuck in the "Pending" state.

Possible workaround for the issue: Although this workaround is not robust, the user can try waiting until all "rm" pods time out. If this does not seem to happen, the user can try to delete the rm pods manually.

The upstream fix has been merged [2]. The fix also introduces test cases.

Here [3] is the PR for back-porting this fix to Che 7.34.x.

[1]https://github.com/eclipse/che/issues/20216
[2]https://github.com/eclipse-che/che-server/pull/79
[3]https://github.com/eclipse-che/che-server/pull/91