[LU-7996] Load Balance After Adding New OST's Created: 07/Apr/16  Updated: 11/Apr/16  Resolved: 11/Apr/16

Status: Resolved
Project: Lustre
Component/s: None
Affects Version/s: Lustre 1.8.9
Fix Version/s: None

Type: Question/Request Priority: Minor
Reporter: Frank (Inactive) Assignee: WC Triage
Resolution: Done Votes: 0
Labels: None
Environment:

RHEL 5.9


Rank (Obsolete): 9223372036854775807

 Description   

I added about 30 TB of space to our nearly full Lustre file system. Now, the original OSTs are almost full with only about 500GB left. How can I even out the space between them and the new OSTs? I was reading on lfs_migrate bit I am unsure of the syntax and if I should run it on the OSS? Thanks.



 Comments   
Comment by Andreas Dilger [ 11/Apr/16 ]

There is a man page for lfs_migrate even in 1.8.9. Note, however, that migration on Lustre 1.8.x is not safe against use/modification by other processes/clients that are using a file during migration. You will somehow have to "know" that files are idle based on how your workload and environment are configured, or which users/applications are currently active. The main concern is if files are being modified during migration and the new data is discarded (despite checks to try and avoid this), or if processes have the old version of the file open and do not see new data being written to the file after it is migrated.If your applications don't modify files after they have been written then this is not a concern.

In order to reduce this risk, you can selectively migrate only older files, for example only large files changed more than 7 days ago:

lfs find /testfs -size +4G -ctime +7d | lfs_migrate

With Lustre 2.5.x and later, it is possible to safely migrate files while they might be open and in use.

Generated at Sat Feb 10 02:13:43 UTC 2024 using Jira 9.4.14#940014-sha1:734e6822bbf0d45eff9af51f82432957f73aa32c.