My guess is that the trouble is caused by the Bytes added or moved while the quotacheck-scan is running. That gives you an estimate of the size of the
inaccuracy.
We gave the users/groups quota in the order of 100 - 1000 Terabytes. During quotacheck, people would not write more than a few Terabytes, so the
effect is negligible. Comparison with 'du' results showed no dramatic deviation.
quotacheck can be painful for the MDS, though. So if you have to go through
several 100 M files - that might be a point for a quiescent system.
Regards,
Thomas
On 03/17/2016 04:38 PM, John White wrote:
So in the lustre manual, it says:
lfs quotacheck requires the file system to be quiescent (i.e. no modifying
operations like write, truncate, create or delete should run
concurrently). Failure to follow this caution may result in inaccurate
user/group disk usage. Operations that do not change Lustre files (such as read
or mount) are okay to run.
With that understood, what sort of inaccuracies are introduced by a live
filesystem (with writers) during a quotacheck? We have a file system we
desperately need to use the quota mechanisms to get even ballpark estimates of
per-user usage. A downtime is not really in the cards at the moment. Further,
the file system isn’t long for this world and will be blown away in a matter of
months (we just need to make it that long).
Will writers on the fs make the data absolutely useless or will it just make
things… fuzzy?
_______________________________________________
lustre-discuss mailing list
[email protected]
http://lists.lustre.org/listinfo.cgi/lustre-discuss-lustre.org
_______________________________________________
lustre-discuss mailing list
[email protected]
http://lists.lustre.org/listinfo.cgi/lustre-discuss-lustre.org
