[Sheepdog] [PATCH 0/2] fix vdi operation bugs

[MORITA Kazutaka]

At Wed, 26 Oct 2011 18:06:17 +0100,
Chris Webb wrote:
> 
> Hi Kazutaka. I pulled these fixes (your devel branch is master + these fixes
> at the moment) and rebuilt. However, I'm afraid I'm still seeing flaky
> operation.
> 
> I do a successful
> 
>   collie vdi create 7fa6adbe-2551-4a60-ab50-c901c972d11d 539545600
> 
> then a bunch of successful setattr, getattr and writes, but after a few
> blocks of writing, end up with
> 
> collie vdi write 7fa6adbe-2551-4a60-ab50-c901c972d11d 16777216
> failed to write object, 2eb90900000004 No object found
> failed to write vdi
> exit 1
> 
> I've put the corresponding sheep.logs at
> 
>   http://cdw.me.uk/tmp/sheep-00.log
>   http://cdw.me.uk/tmp/sheep-01.log
>   http://cdw.me.uk/tmp/sheep-02.log
> 
> After a couple more repeats of the create-and-upload operation, it gets this
> bad:
> 
> 0026# collie vdi list
>   name        id    size    used  shared    creation time   vdi id
> ------------------------------------------------------------------
>   7fa6adbe-2551-4a60-ab50-c901c972d11d     1  515 MB   16 MB  0.0 MB 2011-10-26 16:56   2eb909
> failed to read object, 809a7d7500000000 No object found
> failed to read a inode header
>   9fbb9334-fc73-406d-8641-2cbaedf61d10     1  515 MB  4.0 MB  0.0 MB 2011-10-26 17:00   e7f8bc
> 
> In the case that caused this, collie create worked then setattr -x failed:
> 
> collie vdi create c4053f86-11bf-4a75-808f-37d80ac7b92d 539545600
> exit 0
> 
> collie vdi setattr -x c4053f86-11bf-4a75-808f-37d80ac7b92d lock
> failed to set attr, I/O error
> exit 1
> 
> sheep.logs for this more extended test at
> 
>   http://cdw.me.uk/tmp/sheep2-00.log
>   http://cdw.me.uk/tmp/sheep2-01.log
>   http://cdw.me.uk/tmp/sheep2-02.log

Thanks, I'll dig into this issue again.

> 
> Sorry to be so regularly the bearer of problem reports! I think the big

I really appreciate your feedback. :)

> difference with automated users of sheepdog like our management system is that
> collie calls are executed quite rapidly in close succession, so we create and
> immediate setattr and so on, which might be why we are turning up so many
> timing problems.

I think it's time to introduce a testing framework to Sheepdog so that
developers can reproduce the problems easily.


Thanks,

Kazutaka


> 
> Best wishes,
> 
> Chris.
> -- 
> sheepdog mailing list
> sheepdog at lists.wpkg.org
> http://lists.wpkg.org/mailman/listinfo/sheepdog