[erlang-questions] Problems with Mnesia and failure situations

[Teemu Antti-Poika]

On Thu, Apr 23, 2009 at 10:22 AM, Hakan Mattsson <hakan@REDACTED> wrote:
> On Thu, 23 Apr 2009, Teemu Antti-Poika wrote:
>
>> On Wed, Apr 22, 2009 at 6:56 PM, Hakan Mattsson <hakan@REDACTED> wrote:
>
> Mnesia should reconfigure itself rather quickly when node B is
> shutdown, but the ongoing transactions will have to pay a little
> penalty when this happens. Depending on which phase that the
> transaction is in, different actions are taken.
>
> What is the purpose of shutting down your application if a Mnesia
> transaction takes too long time? Are you trying to implement hard
> real-time behavior on top of Mnesia?

No, we're not going for real-time.

In this case application was trying to react to the failover by
terminating a part of the supervision tree. Since one of the
supervised processes was stuck waiting for a mnesia transaction to
finish, it could not respond to the exit signal.

I'm afraid I no longer have the logs left from the occasion, so I
cannot check the above (why would the termination of the supervisor
itself fail here?). In any case the application recovered via the main
supervisor and that's the important thing.

I may have the time to repeat the experiment tomorrow, but today I
cannot make it.

Teemu A-P