Stop Lustre Filesystem on NXD-Enabled Systems

Considerations when stopping Lustre on an NXD-enabled system.

This procedure describes how to stop the Lustre file system on an NXD-enabled system. NXD caching must be disabled before stopping the file system. The procedure assumes that NXD is already enabled.

Stop all client IO to the system.
Stop NXD caching on the system.
```
admin@cls12345n000$ cscli nxd disable
```
Monitor the NXD caching state by running the cscli nxd list command repeatedly and checking the output.
```
admin@cls12345n000$ cscli nxd list
```
Wait until the value displayed in the Caching State field is disabled for all OSS nodes and OSTs, before proceeding to the next step.
If there is a valid reason not to wait for the caching state to become disabled on all OSS nodes and OSTs, then use the --nxdforce option on the cscli unmount command in the next step. This option could result in data corruption.
Stop Lustre.
```
admin@cls12345n000$ cscli unmount
```
The command will check first to make sure that NXD caching has been disabled on all OSS nodes and OSTs. If not, the command will not unmount the file system. To bypass the check and force the file system to unmount, add the --nxdforce option.
```
admin@cls12345n000$ cscli unmount --nxdforce
```

CAUTION: Using the --nxdforce option could result in data corruption.

Warnings

Do not run a file system integrity check on GridRAID arrays after unmounting Lustre using the --nxdforce option. This is because data in the NXD cache will not be flushed completely.
Do not stop the NXD service after unmounting Lustre using the --nxdforce option.
If faulty or degraded OSTs occur, consult with appropriate support and engineering personnel; then proceed with the recovery process in a controlled environment.