airship/porthole
Code Issues Proposed changes

[ceph]: Added procedure to stop the osd pod from being scheduled

Browse Source 
Change-Id: I7d39f5fdfe9a198baaadfc0f56fbf7b7d0a8fc6b
This commit is contained in:
Venkata, Krishna (kv988c) 2019-08-22 11:39:35 -05:00 committed by KHIYANI, RAHUL (rk0850)
parent 69d9e6db4c
commit 68ea0f9bfa
1 changed files with 21 additions and 6 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

21
docs/ceph_maintenance.md
View File

@ -42,7 +42,21 @@ utilscli osd-maintenance reweight_osds
## 2. Replace failed OSD ##
In the context of a failed drive, Please follow below procedure. Following commands should be run from utility container
In the context of a failed drive, Please follow below procedure.
Disable OSD pod on the host from being rescheduled
kubectl label nodes --all ceph_maintenance_window=inactive
Replace `<NODE>` with the name of the node were the failed osd pods exist.
kubectl label nodes <NODE> --overwrite ceph_maintenance_window=active
Replace `<POD_NAME>` with failed OSD pod name
kubectl patch -n ceph ds <POD_NAME> -p='{"spec":{"template":{"spec":{"nodeSelector":{"ceph-osd":"enabled","ceph_maintenance_window":"inactive"}}}}}'
Following commands should be run from utility container
Capture the failed OSD ID. Check for status `down`
@ -54,9 +68,10 @@ Remove the OSD from Cluster. Replace `<OSD_ID>` with above captured failed OSD I
Remove the failed drive and replace it with a new one without bringing down the node.
Once new drive is placed, delete the concern OSD pod in `error` or `CrashLoopBackOff` state. Replace `<pod_name>` with failed OSD pod name.
Once new drive is placed, change the label and delete the concern OSD pod in `error` or `CrashLoopBackOff` state. Replace `<POD_NAME>` with failed OSD pod name.
kubectl delete pod <pod_name> -n ceph
kubectl label nodes <NODE> --overwrite ceph_maintenance_window=inactive
kubectl delete pod <POD_NAME> -n ceph
Once pod is deleted, kubernetes will re-spin a new pod for the OSD. Once Pod is up, the osd is added to ceph cluster with weight equal to `0`. we need to re-weight the osd.