Xfs delaylog hanged up

[Spelic]

On 11/24/2010 11:52 PM, Spelic wrote:
> On 11/22/2010 08:27 PM, Spelic wrote:
>> ....
>> I was doing a parallel-kernel-unpack benchmark:
>> ....
>
> I tried to reproduce it today but unfortunately I wasn't able to, not 
> even on the same hardware.
>
> The first time it happened so soon after starting such 
> parallel-kernel-unpack benchmark (at the second run!) that I was 
> confident it was possible to reproduce it easily, but today in many 
> hours it never happened again. Sorry. Unfortunately the machine is 
> going back to production soon so I'm not sure I can do further testing.


I had another smaller hangup, and unfortunately I wasn't paying 
attention. I just noticed that between a zillion runs of the "benchmark" 
taking 1 minute (for the part before the sync) I just had one that took 
2m30s for the same part. The system was absolutely idle except that test 
so I can't really see an explanation other than being another smaller 
hangup...

The good news is that it definitely came out of the hangup by itself: 
this time I didn't press Ctrl-C. So even if it was an xfs bug it doesn't 
seem fatal.

The dmesg this time showed nothing; probably because the hangup this 
time was shorter than 120 seconds so the watchdog didn't notice anything.