With the mgmt network reconfiguration, the 'cephfs-nodeplugin'
and 'rbd-nodeplugin' pods don't recognize the change and are
not automatically restarted.
To resolve this, an annotation was added to each
nodeplugin-daemonset.yaml, monitoring any changes to
csiConfig values.
$ kubectl get configmaps -n kube-system cephfs-csiplugin-config -o yaml
apiVersion: v1
data:
config.json: '[{"clusterID":"e36fa26b-bdd5-42a5-83b3-c36a4042109b",
"monitors":["192.168.204.2:6789"]}]'
kind: ConfigMap
...
----------------------------------------------------------------------
$ kubectl get configmaps -n kube-system cephfs-csiplugin-config -o yaml
apiVersion: v1
data:
config.json: '[{"clusterID":"e36fa26b-bdd5-42a5-83b3-c36a4042109b",
"monitors":["192.168.208.3:6789"]}]'
kind: ConfigMap
...
As can be seen above, the monitor is modified after reconfiguration.
This information is obtained through Values.csiConfig in
csiplugin-configmap template:
data:
config.json: |-
{{ toJson .Values.csiConfig | indent 4 -}}
Test Plan:
- PASS: Create pod on AIO-SX with ceph backend
- PASS: Reconfiguration of the mgmt network
- PASS: Check if the previously created pod is running
Story: 2010722
Task: 49295
Change-Id: I9892f1ba7127fe381cba5727e34d16803f93b5bf
Signed-off-by: Erickson Silva de Oliveira <Erickson.SilvadeOliveira@windriver.com>
5255e744eb