Liam,

 If you have another zpool configured somewhere, you could always take a 
snapshot of your mdt and then used send/received to copy that snapshot to 
another zpool.  I helped someone do this one time in order to move the mdt to 
new hardware.

-Rick


On 9/14/22, 6:23 PM, "lustre-discuss on behalf of Liam Forbes via 
lustre-discuss" <[email protected] on behalf of 
[email protected]> wrote:

    Today, in our lustre 2.10.3 filesystem, the MDT ran out of inodes. We are 
using ZFS as the backing filesystem.

    [loforbes@mds02 ~]$ df -i -t lustre
    Filesystem                    Inodes    IUsed IFree IUse% Mounted on
    digdug-meta/lustre2-mgt-mdt 83703636 83703636     0  100% 
/mnt/lustre/local/lustre2-MDT0000

    [loforbes@mds02 ~]$ sudo zpool list -v
    NAME   SIZE  ALLOC   FREE  EXPANDSZ   FRAG    CAP  DEDUP  HEALTH  ALTROOT
    digdug-meta   744G   721G  23.2G         -    86%    96%  1.00x  ONLINE  -
      mirror   372G   368G  4.25G         -    84%    98%
        scsi-35000c5003017156b      -      -      -         -      -      -
        scsi-35000c500301715e7      -      -      -         -      -      -
      mirror   372G   353G  19.0G         -    88%    94%
        scsi-35000c5003017155f      -      -      -         -      -      -
        scsi-35000c500301715a7      -      -      -         -      -      -

    When we try to delete files, we get the error message:
      rm: cannot remove XXXXX: No space left on device

    Is there a way to unlink files and free up inodes?

    Is it possible to expand the existing zpool and filesystem for the MDT?

    Is it possible to do a backup of just our MDT? If so, how?


    -- 
    Regards,
    -liam

    -There are uncountably more irrational fears than rational ones. -P. Dolan
    Liam Forbes           [email protected]                       ph: 
907.450.8618
    UAF GI Research Computing Systems Manager
    hxxps://calendly.com/ualoforbes/30min

_______________________________________________
lustre-discuss mailing list
[email protected]
http://lists.lustre.org/listinfo.cgi/lustre-discuss-lustre.org

Reply via email to