[LU-10917] If setstripe for DoM file fails, less/no information is given Created: 16/Apr/18 Updated: 17/Apr/18 Resolved: 17/Apr/18 |
|
| Status: | Resolved |
| Project: | Lustre |
| Component/s: | None |
| Affects Version/s: | None |
| Fix Version/s: | None |
| Type: | Bug | Priority: | Minor |
| Reporter: | Rahul Deshmukh (Inactive) | Assignee: | Mikhail Pershin |
| Resolution: | Not a Bug | Votes: | 0 |
| Labels: | None | ||
| Issue Links: |
|
||||||||
| Severity: | 3 | ||||||||
| Rank (Obsolete): | 9223372036854775807 | ||||||||
| Description |
|
While exploring DoM feature, found out that the error returned by setstripe on dom file As a work around, can we change this CDEBUG to CWARN (also for other similar error /* Check stripe size againts per-MDT limit */ if (stripe_size > d->lod_dom_max_stripesize) { CDEBUG(D_LAYOUT, "DoM component size " "%u is bigger than MDT limit %u, check " "dom_max_stripesize parameter\n", stripe_size, d->lod_dom_max_stripesize); RETURN(-EINVAL); } Please suggestion, if this is ok, will create patch and push. |
| Comments |
| Comment by Peter Jones [ 16/Apr/18 ] |
|
Mike Could you please advise on this one? Thanks Peter |
| Comment by Andreas Dilger [ 17/Apr/18 ] |
|
Most users will not be able to see console messages on the MDS, so this will not really help them. It would be better to print a useful error message at the client from "lfs", iff there was an error returned from the MDS. Note that doing "pre-checking" of all the parameters on the client can be problematic, since this may cause failures if the server-side limits change. That said, there is work underway to make the MDS more "accepting" of the -L mdt component being too large. Since it is difficult for the user to know what the dom_stripesize limit is (in the future we may want to reduce this dynamically if the MDS space is running short), and it may be different for each MDT, then it is better if the MDS adjusts the mdt component size to match its current limit, rather than returning an error. This is being worked on under |
| Comment by Rahul Deshmukh (Inactive) [ 17/Apr/18 ] |
|
Thanks Andreas, we can close this issue. |
| Comment by Peter Jones [ 17/Apr/18 ] |
|
ok - thanks Rahul |