[Pacemaker] Pacemaker Digest, Vol 30, Issue 40

[Aleksey Zholdak]

>>> The sbd daemon needs to be running on both nodes (the openais init script
>>> should take care of that on SLES), but there only needs to be one sbd
>>> primitive, it does not need to be cloned.  Pacemaker will make sure it
>>> is running somewhere, which is enough.
>>
>> What you tell about? sbd must be running on _each_ node, not "somewhere"!
>
> Think of "sbd must be running on _each_ node" as
> "network power switch port needs to be accessible from every where",
> and of "sbd _primitive_" (stonith resource agent) "running somewhere"
> as "the telnet that will tell the power switch to reset some node".
> One telnet is enough, as long as it can reach the power switch.
> And several telnets may interfere, and likely report failure,
> if the power switch only supports one connection at a time.
> (Which, in a way, is true for SBD daemon).
> So please: do not clone SBD.

As I understand an sbd theory each node periodically writes its status to 
own slot on shared partition.
I configure as recommended and its works fine and clear.
I can not agree that you are right and this topic describes another trouble.