On Wed, Oct 29, 2014 at 05:46:17PM +0800, Zeng Linggang wrote: > Hello, > > When I take XFS filesystem as test target, the case generic/233 often > fails in my test machines (RHEL7.0GA: 3.10.0-123) with output like this: > generic/233 [failed, exit status 1] - output mismatch (see > /var/lib/xfstests/results//generic/233.out.bad) > --- tests/generic/233.out 2014-10-22 14:17:06.239461640 +0800 > +++ results/generic/233.out.bad 2014-10-23 14:39:48.210480319 +0800 > @@ -4,4 +4,7 @@ > seed = S > Comparing user usage > -Comparing group usage > +4c4 > +< #1001 -- 31348 32000 32000 772 1000 1000 > +--- > +> #1001 -- 31144 32000 32000 772 1000 1000 > ... > generic/231 and generic/270 also have the similar quota accounting mismatch. fsstress can result in different file and directory layouts, hence result in different amounts of metadata in the filesystem. Metadata is also accounted to quotas (e.g. to the user who owns the directory) and so even if data usage is identical the metadata can still vary. I see these test give small variations quite frequently, but they are are usually only quite small (a handful of blocks) so I've never really worried about it too much. Cheers, Dave. -- Dave Chinner david@xxxxxxxxxxxxx _______________________________________________ xfs mailing list xfs@xxxxxxxxxxx http://oss.sgi.com/mailman/listinfo/xfs
