[Pacemaker] [Partially SOLVED] pacemaker/dlm problems

[Andrew Beekhof]

On Tue, Jan 17, 2012 at 12:02 PM, Andrew Beekhof <andrew at beekhof.net> wrote:
> On Mon, Jan 16, 2012 at 5:49 PM, Vladislav Bogdanov
> <bubble at hoster-ok.com> wrote:
>> 16.01.2012 09:17, Andrew Beekhof wrote:
>> [snip]
>>>> At the same time, stonith_admin -B succeeds.
>>>> The main difference I see is st_opt_sync_call in a latter case.
>>>> Will try to experiment with it.
>>>
>>> /Shouldn't/ matter.
>>
>> It really looks like it matters.
>
> It does at least give you the real result rather than a call-id.
> But even without it you should be getting:
>            log_error("Unknown result when kicking node %d from the
> cluster", nodeid);
>
> or at least some of the logs prior to it.

Can you try replacing

int rc = pacemaker_terminate_member(nodeid);

with:

int rc = stonith_api_cs_kick(nodeid, 0, TRUE);

"TRUE" because if the dlm is asking for fencing, the node probably
shouldn't come back on its own.
It also uses st_opt_sync_call.