Document about stale resource cleanup

1. when user delete pv manual, it will result in stale metadata and image in ceph
This commit is contained in:
chenxu1990 2020-04-03 15:10:29 +08:00 committed by mergify[bot]
parent 34fc1d847e
commit 5475022bc3
2 changed files with 100 additions and 0 deletions

View File

@ -36,6 +36,7 @@ Independent CSI plugins are provided to support RBD and CephFS backed volumes,
for CephFS plugin configuration and deployment please for CephFS plugin configuration and deployment please
refer [cephfs doc](https://github.com/ceph/ceph-csi/blob/master/docs/deploy-cephfs.md). refer [cephfs doc](https://github.com/ceph/ceph-csi/blob/master/docs/deploy-cephfs.md).
- For example usage of RBD and CephFS CSI plugins, see examples in `examples/`. - For example usage of RBD and CephFS CSI plugins, see examples in `examples/`.
- Stale resource cleanup, please refer [cleanup doc](docs/resource-cleanup.md).
NOTE: NOTE:

99
docs/resource-cleanup.md Normal file
View File

@ -0,0 +1,99 @@
# Stale Resource Cleanup
If the PVC is created with storage class which is having the `reclaimPolicy`
as `Retain` will not delete the PV object, backend omap metadata and backend image.
Manual deletion of PV will result in stale omap keys, values,
cephfs subvolume and rbd image.
It is required to cleanup metadata and image separately.
## Steps
### 1. Get PV name from PVC
a. get pv_name
`[$]kubectl get pvc pvc_name -n namespace -owide`
```bash
$ kubectl get pvc mysql-pvc -owide -n prometheus
NAME STATUS VOLUME
mysql-pvc Bound pvc-bc537af8-67fc-4963-99c4-f40b3401686a
CAPACITY ACCESS MODES STORAGECLASS AGE VOLUMEMODE
20Gi RWO csi-rbd 14d Filesystem
```
### 2. Get omap key/value
a. get omapkey (suffix of csi.volumes.default is value used for the CLI option
[--instanceid](deploy-rbd.md#configuration) in the provisioner deployment.)
`[$]rados listomapkeys csi.volumes.default -p pool_name | grep pv_name`
```bash
$ rados listomapkeys csi.volumes.default -p kube_csi | grep pvc-bc537af8-67fc-4963-99c4-f40b3401686a
csi.volume.pvc-bc537af8-67fc-4963-99c4-f40b3401686a
```
b. get omapval
`[$]rados getomapval csi.volumes.default omapkey -p pool_name`
```bash
$ rados getomapval csi.volumes.default csi.volume.pvc-bc537af8-67fc-4963-99c4-f40b3401686a
-p kube_csi
value (36 bytes) :
00000000 64 64 32 34 37 33 64 30 2d 36 61 38 63 2d 31 31 |dd2473d0-6a8c-11|
00000010 65 61 2d 39 31 31 33 2d 30 61 64 35 39 64 39 39 |ea-9113-0ad59d99|
00000020 35 63 65 37 |5ce7|
00000024
```
### 3. Delete the RBD image or CephFS subvolume
a. remove rbd image(csi-vol-omapval, the prefix csi-vol is value of [volumeNamePrefix](deploy-rbd.md#configuration))
`[$]rbd remove rbd_image_name -p pool_name`
```bash
$ rbd remove csi-vol-dd2473d0-6a8c-11ea-9113-0ad59d995ce7 -p kube_csi
Removing image: 100% complete...done.
```
b. remove cephfs subvolume(csi-vol-omapval)
`[$]ceph fs subvolume rm volume_name subvolume_name group_name`
```bash
$ ceph fs subvolume rm cephfs csi-vol-340daf84-5e8f-11ea-8560-6e87b41d7a6e csi
```
### 4. Delete omap object and omapkey
a. delete omap object
`[$]rados rm csi.volume.omapval -p pool_name`
```bash
$ rados rm csi.volume.dd2473d0-6a8c-11ea-9113-0ad59d995ce7 -p kube_csi
```
b. delete omapkey
`[$]rados rmomapkey csi.volumes.default csi.volume.omapkey -p pool_name`
```
$ rados rmomapkey csi.volumes.default csi.volume.pvc-bc537af8-67fc-4963-99c4-f40b3401686a
-p kube_csi
```
### 5. Delete PV
a. delete pv
`[$] kubectl delete pv pv_name -n namespace`
```bash
$ kubectl delete pv pvc-bc537af8-67fc-4963-99c4-f40b3401686a -n prometheus
persistentvolume "pvc-bc537af8-67fc-4963-99c4-f40b3401686a" deleted
```