Hello all,
I have a weird behavior on my Cephfs.
On May 29th I noticed a drop of 50Tb in my data pool. It has been followed
by a decrease of space usage in the metadata pool since then.
From May 29th, still happening as I write, the metadata pool has lost 1Tb
over the initial 1.8Tb.
Regarding the number of objects, it was 8.4 Million and is now 7.8 Million.
I assume that my users have deleted a lot a of files on that day (our
cephfs consists a very small files of about 4Mb).
But such a huge decrease in metadata pool got me really concerned, it
really seems huge.
I thought that maybe it was the mds lazy deletion happening but I am not
sure.
Does anyone have any thought about this ?
Do you know more about lazy deletion ? There is not much documentation
about this online.
Do you know any way, command, logfile to search in order to get the current
lazy deletion operations ?
I noticed that the read_ops, read_bytes, write_ops and write_bytes reported
by the command rados df detail are negative for the metadata pool.
My cluster is running Nautilus.
Any help would be appreciated,
Best regards,
Nate
Show replies by date