[sheepdog] [PATCH 1/2] test: use corosync as default

Liu Yuan namei.unix at gmail.com
Mon Aug 20 07:10:25 CEST 2012


On 08/20/2012 12:31 PM, MORITA Kazutaka wrote:
> At Mon, 20 Aug 2012 10:42:24 +0800,
> Liu Yuan wrote:
>>
>> On 08/20/2012 10:15 AM, MORITA Kazutaka wrote:
>>> On my environment, both corosync and local drivers cannot pass the
>>> test (026 on the current master).  Seems that it is not a local driver
>>> specific problem.
>>
>> Corosync can pass it on my env, I guess that is the problem of the test
>> script.(You can manually issue collie vdi list to see if there are 40
>> images created)
>>
>> Local driver can't pass it because it doesn't support confchg priority
>> over notify event.
> 
> I got something like the following outputs with both drivers:
> 
> ==
> 026      - output mismatch (see 026.out.bad)
> --- 026.out     2012-08-20 13:20:39.000000000 +0900
> +++ 026.out.bad 2012-08-20 13:22:28.000000000 +0900
> @@ -1,3 +1,7 @@
>  QA output created by 026
>  using backend farm store
> -42
> +Failed to read object 80fd366200000000 No object found
> +Failed to read inode header
> +Failed to read object 80fd366200000000 No object found
> +Failed to read inode header
> +41
> Failures: 026
> Failed 1 of 1 tests
> ==
> 
> It seems that I'm hitting a different problem.
> 

Yes, but it might be the problem of script. We kill 5 node in one go,
which we are risking to lose object in some cases. I'll come up with a
fix to this script.

Thanks,
Yuan



More information about the sheepdog mailing list