Re: [PATCH 4/8] Add test of quota accounting using fsx

Re: [PATCH 4/8] Add test of quota accounting using fsx

[Eric Sandeen]

On 06/15/2010 04:55 AM, Jan Kara wrote:
> On Wed 09-06-10 12:49:49, Eric Sandeen wrote:
>> Jan Kara wrote:
>>> Run fsx (and also several fsx threads in parallel) and verify that
>>> quota accounting is correct after they finish.
>>
>> Jan, I'm having trouble with this one on XFS for some reason, with our 
>> RHEL6 kernel and quota-3.17...
>   OK, attached is an improvement to the XFSQA tests after which all quota
> tests pass for XFS just fine.
>   The second patch is just minor general improvement of _require_scratch
> macro.
>   Could they be added to XFSQA repository? Thanks.

Jan, I've got some ext4 failures reported on these, although I can't hit
them, so not quite sure what's going on.

In 231:

+< fsgqa     --     760       0       0              3     0     0
+---
+> fsgqa     --     764       0       0              3     0     0
+14c14
+< fsgqa     --     760       0       0              3     0     0
+---
+> fsgqa     --     764       0       0              3     0     0

after the quotacheck & repquota we have 4 more blocks.  Maybe this
is due to my accounting of metadata blocks at write time, and not
before ... would it be reasonable to put a sync call as the first
line of check_usage() ?

Also in 233:

+< #501      --   15392       0       0            998     0     0
+< #501      --   15392   32000   32000            998  1000  1000
+---
+> #501      +-   32084   32000   32000  7days     998  1000  1000
+> #501      --   32084       0       0            998     0     0

"7days" magically appeared after the quotacheck.  Not sure what's going
on there...

Thanks,
-Eric

Re: [PATCH 4/8] Add test of quota accounting using fsx

[Jan Kara]

On Mon 26-07-10 16:46:17, Eric Sandeen wrote:
> On 06/15/2010 04:55 AM, Jan Kara wrote:
> > On Wed 09-06-10 12:49:49, Eric Sandeen wrote:
> >> Jan Kara wrote:
> >>> Run fsx (and also several fsx threads in parallel) and verify that
> >>> quota accounting is correct after they finish.
> >>
> >> Jan, I'm having trouble with this one on XFS for some reason, with our 
> >> RHEL6 kernel and quota-3.17...
> >   OK, attached is an improvement to the XFSQA tests after which all quota
> > tests pass for XFS just fine.
> >   The second patch is just minor general improvement of _require_scratch
> > macro.
> >   Could they be added to XFSQA repository? Thanks.
> 
> Jan, I've got some ext4 failures reported on these, although I can't hit
> them, so not quite sure what's going on.
> 
> In 231:
> 
> +< fsgqa     --     760       0       0              3     0     0
> +---
> +> fsgqa     --     764       0       0              3     0     0
> +14c14
> +< fsgqa     --     760       0       0              3     0     0
> +---
> +> fsgqa     --     764       0       0              3     0     0
> 
> after the quotacheck & repquota we have 4 more blocks.  Maybe this
> is due to my accounting of metadata blocks at write time, and not
> before ... would it be reasonable to put a sync call as the first
> line of check_usage() ?
  Just last week a change went into xfstests which introduces a generic
quota checking function and uses sync before getting quota usage. I think
xfstests passed for me with ext4 after this change but I've now restarted
the tests to recheck it.
  
> Also in 233:
> 
> +< #501      --   15392       0       0            998     0     0
> +< #501      --   15392   32000   32000            998  1000  1000
> +---
> +> #501      +-   32084   32000   32000  7days     998  1000  1000
> +> #501      --   32084       0       0            998     0     0
> 
> "7days" magically appeared after the quotacheck.  Not sure what's going
> on there...
  That's because the usage after checking exceeded block soft limit and
thus grace time has been set. So it's the same problem as in the above
test.

								Honza
-- 
Jan Kara <jack@suse.cz>
SUSE Labs, CR

Re: [PATCH 4/8] Add test of quota accounting using fsx

[Jan Kara]

On Tue 27-07-10 10:15:38, Jan Kara wrote:
> On Mon 26-07-10 16:46:17, Eric Sandeen wrote:
> > On 06/15/2010 04:55 AM, Jan Kara wrote:
> > > On Wed 09-06-10 12:49:49, Eric Sandeen wrote:
> > >> Jan Kara wrote:
> > >>> Run fsx (and also several fsx threads in parallel) and verify that
> > >>> quota accounting is correct after they finish.
> > >>
> > >> Jan, I'm having trouble with this one on XFS for some reason, with our 
> > >> RHEL6 kernel and quota-3.17...
> > >   OK, attached is an improvement to the XFSQA tests after which all quota
> > > tests pass for XFS just fine.
> > >   The second patch is just minor general improvement of _require_scratch
> > > macro.
> > >   Could they be added to XFSQA repository? Thanks.
> > 
> > Jan, I've got some ext4 failures reported on these, although I can't hit
> > them, so not quite sure what's going on.
> > 
> > In 231:
> > 
> > +< fsgqa     --     760       0       0              3     0     0
> > +---
> > +> fsgqa     --     764       0       0              3     0     0
> > +14c14
> > +< fsgqa     --     760       0       0              3     0     0
> > +---
> > +> fsgqa     --     764       0       0              3     0     0
> > 
> > after the quotacheck & repquota we have 4 more blocks.  Maybe this
> > is due to my accounting of metadata blocks at write time, and not
> > before ... would it be reasonable to put a sync call as the first
> > line of check_usage() ?
>   Just last week a change went into xfstests which introduces a generic
> quota checking function and uses sync before getting quota usage. I think
> xfstests passed for me with ext4 after this change but I've now restarted
> the tests to recheck it.
  For me all the quota tests pass just fine with ext4 and the latest
xfstests... So does the latest version work also for you?

									Honza
-- 
Jan Kara <jack@suse.cz>
SUSE Labs, CR

Re: [PATCH 4/8] Add test of quota accounting using fsx

[Eric Sandeen]

On Jul 27, 2010, at 3:48 AM, Jan Kara <jack@suse.cz> wrote:

>>> 
>>  Just last week a change went into xfstests which introduces a generic
>> quota checking function and uses sync before getting quota usage. I think
>> xfstests passed for me with ext4 after this change but I've now restarted
>> the tests to recheck it.
>  For me all the quota tests pass just fine with ext4 and the latest
> xfstests... So does the latest version work also for you?
> 
Whoops sorry, they do pass for me in the devel tree; I had missed that recent change and I guess our qa had tested a tree without it.  I wondered why it didn't work for them...  Sorry for the noise....

Eric

>                                    Honza
> --
> Jan Kara <jack@suse.cz>
> SUSE Labs, CR
>