Re: Change in integration test for javelin

[Rohit Yadav <bh...@apache.org>]

On Thu, Jan 31, 2013 at 6:22 PM, Prasanna Santhanam <ts...@apache.org> wrote:
> Reverted (1b922e8) the earlier fix for a better commit later.

Prasanna, in the smoke test for vm expunge, we should remove the hard
coded 20 mins as well?

For the revert, Anthony won't this revert reintroduce the thread/deadlock issue?

Regards.

>
> On Sat, Jan 26, 2013 at 05:18:04PM +0530, Rohit Yadav wrote:
>> I don't know why it exists or it has to be this way either, we should
>> just fix it.
>>
>> Anthony, if we check, lock db entity and expunge an entity would that
>> save us from race condition, will that work? In such an approach, will
>> this have any side effects: if I just return from actually expunging
>> an entity which is already expunged?
>>
>> Regards.
>>
>> On Fri, Jan 25, 2013 at 8:53 PM, Anthony Xu <Xu...@citrix.com> wrote:
>> > That's a bad/quick fix for http://bugs.cloudstack.org/browse/CS-15922
>> >
>> > Delete account will expunge all VMs for this account immediate ,
>> > if expunge.interval is too short , VM GC may kick in, then two
>> > threads are trying to expunge the same VMs, which causes expunge
>> > fail.
>> >
>> > I think you can remove the hacker, and reopen the bug.
>> >
>> >
>> > Anthony
>> >
>> >> -----Original Message-----
>> >> From: Min Chen [mailto:min.chen@citrix.com]
>> >> Sent: Friday, January 25, 2013 8:30 PM
>> >> To: cloudstack-dev@incubator.apache.org
>> >> Cc: cloudstack-dev@incubator.apache.org
>> >> Subject: Re: Change in integration test for javelin
>> >>
>> >> Anthony can comment more on, he checked in that code to avoid some race
>> >> conditions in expunge VM.
>> >>
>> >> -min
>> >>
>> >> Sent from my iPhone
>> >>
>> >> On Jan 25, 2013, at 7:59 PM, "David Nalley" <da...@gnsa.us> wrote:
>> >>
>> >> > On Fri, Jan 25, 2013 at 9:57 PM, prasanna <ts...@apache.org> wrote:
>> >> >> I know this code exists and it fails all the expunge tests
>> >> internally.
>> >> >> Do you know why it exists? Why should expunge be > 600 always? Seems
>> >> >> like we're overriding the global setting that the admin sets and
>> >> >> without log/warning.
>> >> >
>> >> > I concur - no idea why we would do this. If for no other reason than
>> >> > it makes testing too long, sounds like a bad idea.
>> >> >
>> >> > --David
>
> --
> Prasanna.,

Re: Change in integration test for javelin

[prasanna <ts...@apache.org>]

On 25 February 2013 07:43, Rohit Yadav <bh...@apache.org> wrote:
> On Thu, Jan 31, 2013 at 6:22 PM, Prasanna Santhanam <ts...@apache.org> wrote:
>> Reverted (1b922e8) the earlier fix for a better commit later.
>
> Prasanna, in the smoke test for vm expunge, we should remove the hard
> coded 20 mins as well?
>
> For the revert, Anthony won't this revert reintroduce the thread/deadlock issue?

Guess that came in from the javelin merge. fixed now (f5ba913)