[lustre-discuss] lfs migrate status

[Michael DiDomenico]

the servers are 2.12 and the client is 2.15.  yes it was single
thread.  it does still seem to be running in the background somewhere
as i see directories switch out of none,migratiing state over time.
i hope lsfck isn't the only fix, i can't exactly unmount the
filesystem at the moment.

On Wed, Jun 19, 2024 at 11:47 AM Alex Zarochentsev <zamenator at gmail.com> wrote:
>
> Hello,
>
> On Mon, Jun 17, 2024 at 4:04 PM Michael DiDomenico via lustre-discuss <lustre-discuss at lists.lustre.org> wrote:
>>
>> i am attempting to use lfs migrate to move some directories from one
>> mdt to another.  the lfs migrate part seemed to be okay as i didn't
>> get any errors, but there are directories still in "migrating" state
>>
>> lfs getdirstripe .
>> lmv_stripe_count: 2 lmv_stripe_offset: 1 lmv_hash_type: non,migrating
>> mdtidx
>> 1
>> 0
>
>
> i.e. lfs migrate completed but some dirs are in "none, migrating" state? I think the migration process failed.
> I believe https://jira.whamcloud.com/browse/LU-17635 is going to fix that.
>
> Can you please share info about Lustre versions on client and mds servers and was it just a single thread
> 'lfs migrate' ? Ideally I would like to know how to reproduce such a situation.
>
>>
>> if i do a `mkdir tmp` in the directory currently i get "bad file
>> descriptor".  i figured this process would clean up over the weekend,
>> but it doesn't seem to have finished up
>
>
> setting the "BAD_HASH" flag by LFSCK would allow some operations over those dirs,
> resuming  the migration process or removal of the directories.
>
> Thanks,
> Zam.
>
>>
>>
>> is there a way to see what lustre is doing? or estimate how long it'll
>> take to finish whatever it's doing?  or how to cancel the migrations
>> still outstanding?
>> _______________________________________________
>> lustre-discuss mailing list
>> lustre-discuss at lists.lustre.org
>> http://lists.lustre.org/listinfo.cgi/lustre-discuss-lustre.org