[LU-19783] many fids with "fid2path: No data available"

• Type: Bug
• Resolution: Unresolved
• Priority: Critical
• Affects Version/s: Lustre 2.15.7
• Labels: llnl topllnl
• Environment: TOSS 4.8-18 ; RHEL kernel 4.18.0-553.83 ; lustre 2.15.7_2.llnl https://github.com/llnl/lustre/tree/2.15.7_2.llnl ; zfs 2.2.8
• Severity: 3
• Assignee: Hongchao Zhang
• Reporter: Gian-Carlo Defazio

Description

We have users complaining about having much higher inode counts in "lfs quota" than can be found with "find".

For one user, I found all the fids using zdb.

Of about 155,000,000 fids, only about 350,000 have a path per fid2path. This is consistent with what find,mpifileutils,starfish report for that user.

The rest exist at <fs>/.lustre/.fid/<fid>, but fail fid2path with:

"fs fid2path: cannot find <fs> <fid>: No data available" and return code 61.

These files with no paths are all empty, all on the same MDT (as expected), and were created in a 6.5 hour period.

Is this a normal sort of error that can be addressed with lfsck?

I'm trying to figure out how these files got created and if they can safely be deleted. Also, are there tools that could help find other such files? Right now my plan is to grab the fids of the whole file system and do fid2path for them all.