[LU-4742] bad FID format - should be [0x200000400:0x1:0x0] Created: 10/Mar/14 Updated: 18/Apr/14 Resolved: 18/Apr/14 |
|
| Status: | Resolved |
| Project: | Lustre |
| Component/s: | None |
| Affects Version/s: | Lustre 2.4.1, Lustre 2.6.0 |
| Fix Version/s: | Lustre 2.6.0 |
| Type: | Bug | Priority: | Minor |
| Reporter: | Daire Byrne (Inactive) | Assignee: | Niu Yawei (Inactive) |
| Resolution: | Fixed | Votes: | 0 |
| Labels: | None | ||
| Severity: | 3 |
| Rank (Obsolete): | 13035 |
| Description |
|
I have been seeing this a few times recently. It doesn't seem to be particularly critical but I don't think this file exists on the filesystem. Mar 10 10:29:37 cmds1 kernel: Lustre: 35876:0:(mdd_device.c:563:obf_lookup()) charlie-MDD0000: bad FID format [.fungus_skip], should be [0x200000400:0x1:0x0] Mar 10 10:38:03 cmds1 kernel: Lustre: 35876:0:(mdd_device.c:563:obf_lookup()) charlie-MDD0000: bad FID format [.fungus_skip], should be [0x200000400:0x1:0x0] Mar 10 10:49:20 cmds1 kernel: Lustre: 5100:0:(mdd_device.c:563:obf_lookup()) charlie-MDD0000: bad FID format [.fungus_skip], should be [0x200000400:0x1:0x0] Mar 10 10:59:45 cmds1 kernel: Lustre: 35652:0:(mdd_device.c:563:obf_lookup()) charlie-MDD0000: bad FID format [.fungus_skip], should be [0x200000400:0x1:0x0] Mar 10 11:09:45 cmds1 kernel: Lustre: 35652:0:(mdd_device.c:563:obf_lookup()) charlie-MDD0000: bad FID format [.fungus_skip], should be [0x200000400:0x1:0x0] Mar 10 11:21:01 cmds1 kernel: Lustre: 5100:0:(mdd_device.c:563:obf_lookup()) charlie-MDD0000: bad FID format [.fungus_skip], should be [0x200000400:0x1:0x0] The ".fungus_skip" file is just a blank file that our filesystem trawler tries to write to dirs we know we don't need to scan (OLD/ARCHIVE etc). We do not see this message on a similar (same version) filesystem. The only thing that I think may have happened to this filesystem is that the mds-survey benchmark was aborted halfway through and left some "test" dirs that had to be manually deleted by mounting with ldiskfs. I'm not sure if this has messed up the FIDs. |
| Comments |
| Comment by Peter Jones [ 10/Mar/14 ] |
|
Niu Could you please advise on this one? Thanks Peter |
| Comment by Niu Yawei (Inactive) [ 11/Mar/14 ] |
|
Did you accidently create .fungus_skip in the .lustre directory (.lustre/fid/) when you mount the mds with ldiskfs? |
| Comment by Daire Byrne (Inactive) [ 11/Mar/14 ] |
|
Well no, I don't see how I could have done that. I can remount ldiskfs and remove the file if that helps. Like I said the file is not important. |
| Comment by Andreas Dilger [ 11/Mar/14 ] |
|
I think it is your crawler trying to create this file in the .lustre/fid directory that is causing this error. You should exclude this directory from your crawling. |
| Comment by Andreas Dilger [ 11/Mar/14 ] |
|
The main question is whether this error message is useful at all? It could be improved to indicate this is caused by opening files in .lustre/fid/, or just removed entirely. |
| Comment by Niu Yawei (Inactive) [ 12/Mar/14 ] |
Giving a warnning to inform user what's wrong could be better than return error silently, maybe we just need to re-stating the message to indicate that's just a lookup in .lustre/fid/. |
| Comment by Niu Yawei (Inactive) [ 12/Mar/14 ] |
|
improved warning message: http://review.whamcloud.com/9594 |
| Comment by Daire Byrne (Inactive) [ 12/Mar/14 ] |
|
Okay it looks like our trawler was trying to scan .lustre (over NFS). We have excluded it now and predictably the (confusing!) error message has disappeared. Cheers! |
| Comment by Peter Jones [ 18/Apr/14 ] |
|
Landed for 2.6 |