massive size of ".glusterfs/unlink" underlying fs?

[lejeczek]

Hi guys.
Volume seems healthy - have not parse logs diligently yet - and glusterfs does not complain about anything explicitly, yet...
a volume which has almost nothing on it - glusterfs backend shows - in this case:
1.2T /devs/00.GLUSTERs/PEOPLE/.glusterfs/unlink
Has anybody seen this/similar and if yes.. what is that?
many thanks, L.

[gilbertoferreira]

I am hope you got some answer.
It's been quite difficult nowadays.

[lejeczek]

My gluster cluster was in some - most likely inreproducible circumstances - 3 nodes, one was hard-crashing with underlying fs corruption, second one was had / at 100% for most part of the day and the third node which had that massive unlink subdir was trying, eventually, to heal I think - that must have been one of those "edge cases", which was not possible for GF to handle well, but !... backups ! so no data loss.

[gilbertoferreira]

I would started to stop everything after do a reliable backup.
Then, I'd try to figure out what happens to disk that given you fs corruption.
Or get a set of know good disk and make the cluster again.
Which version are you using, btw?

[lejeczek]

I'm on 11.2 - these thing happen in lab env running experiments - my hard-crashing system was actually having hardware issue, memory specifically, ram settings vs CPU.