[LU-15418] I need clarification on using lctl replace_nids Created: 06/Jan/22 Updated: 04/Aug/22 |
|
| Status: | Open |
| Project: | Lustre |
| Component/s: | None |
| Affects Version/s: | Lustre 2.12.8 |
| Fix Version/s: | None |
| Type: | Question/Request | Priority: | Blocker |
| Reporter: | Roger Sersted | Assignee: | Artem Blagodarenko |
| Resolution: | Unresolved | Votes: | 0 |
| Labels: | lctl | ||
| Environment: |
Clients 2.12.7 - CentOS 7.9 |
||
| Issue Links: |
|
||||||||
| Epic/Theme: | lnet | ||||||||
| Rank (Obsolete): | 9223372036854775807 | ||||||||
| Description |
|
I'm moving some OSTs to a different server within the same subnet as my existing Lustre servers, so the module configs won't change. However, the man page replace_nids is really vague. Given the following: lctl list_nids #current server 172.17.1.104@o2ib Current failover NIDs 172.17.1.105@o2ib And the new OSS 172.17.1.236@o2ib New Failover OSS 172.17.1.237@o2ib The device would be lustrefc-OST000e Would this be the correct command lctl replace_nids lustrefc-OST000e \ 172.17.1.236@o2ib,172.16.1.236@tcp:172.17.1.237@o2ib,172.16.1.237@tcp The above command would be run on the MGS server and the MGT is the only service mounted/running. I'll be moving about 8 OSTs. I'm guessing I should run my commands, then unmount the MGS. Then start Lustre in this order, MGT, MDT then OSTs in ascending order. Correct? The filesystem is down for maintenance. This is my final step before bringing it online. Thank you. |
| Comments |
| Comment by Andreas Dilger [ 07/Jan/22 ] |
|
There is some brief description of the replace_nids command in LU-11846, and a request to the author of that command to provide a separate lctl-replace_nids.8 man page that has a more complete description of that command, and preferably with some usage examples. |
| Comment by Roger Sersted [ 12/Jan/22 ] |
|
I successfully used the writeconf method to move the OSTs to different servers. |