[sheepdog] [PATCH 1/2] test: consoliate test 035 to test more bugs

[MORITA Kazutaka]

At Tue, 04 Sep 2012 16:30:58 +0800,
Liu Yuan wrote:
> 
> On 09/04/2012 12:02 PM, MORITA Kazutaka wrote:
> > Should we call _wait_for_sheep_recovery after _simulate_machine_down?
> > I've encountered no object error with this patch.
> 
> Nope, actually it might be the collie who use modified recv timer and
> didn't write objects successfully into sheep ( I have met this problem
> when I add recv timer). For a sound implementation, we won't have
> problems because gateway will retry request until success.

But the already stored objects can be lost, no?  Your script kills 4
nodes without waiting for recovery completion, though the redundancy
level is 3.

Thanks,

Kazutaka