Multus DaemonSet should be restarted during cidr expansion
Affects | Status | Importance | Assigned to | Milestone | |
---|---|---|---|---|---|
Multus Charm |
Triaged
|
Medium
|
Unassigned |
Bug Description
test_service_
This test attempts to expand the cidr range by updating config service-cidr the kubernetes-
Effect:
New pods scheduled after the change, will fail to contact the API service because it's address will have changed and cannot proceed out of the Pending State.
(eg -- 10.152.183.1 will no longer be the API service, but rather 10.152.182.1)
Workaround:
Manually restart the Daemonset `kube-multus-ds` in the `kube-system` namespace
* This can hang as well -- since the container install-
* In order to replace those binaries, run `killall multus` and `rm /opt/cni/
[1]: https:/
[2]: https:/
Changed in charm-multus: | |
importance: | Undecided → Medium |
status: | New → Triaged |
Skip the cidr expansion test [1]
[1] https:/ /github. com/charmed- kubernetes/ jenkins/ pull/1327