Carson Gaspar wrote:
> Xinwei Hu wrote:
> 
>>   The known issue is that I don't know how to daemonize in bash,
>> so the pingd RA needs a little tweak also.
> 
> You can't daemonize in bash, unless your OS comes with some executable
> that daemonizes arbitrary programs (I think some flavours of Linux do).

Actually, there isn't that much to daemonizing things.  A little I/O
redirection, a little signal handling and a little process group
manipulation (not required for LRM RAs).

There are some disadvantages to this approach - mainly that the same
ping status can't be used for both quorum and setting variables in the
CIB.  And, you can't forge ping packets which might cause misbehavior of
the system because the current ping packets are digitally signed.

Currently we only use the ping status for the CIB function, but that's
something that should change.  In fact, quorum computation was why the
ping function was originally added.

The fact that it was incredibly easy to do also contributed to adding it
long before we had code that used it.

In addition, pingd can use ping_group "members" as well as ping "members".

On the other hand, with this approach, it is as simple as changing a
resource definition to change what's being monitored - which is nice.
Having it in ha.cf is currently messy to maintain.  That's too bad, and
ought to be changed, but it's not likely to happen soon.

What really should be done in general is for the CIB to be able to
replace ha.cf - for ping configuration and several other reasons.  But
that would require enhancements to the CIB infrastructure.

-- 
    Alan Robertson <[EMAIL PROTECTED]>

"Openness is the foundation and preservative of friendship...  Let me
claim from you at all times your undisguised opinions." - William
Wilberforce
_______________________________________________
Linux-HA mailing list
[email protected]
http://lists.linux-ha.org/mailman/listinfo/linux-ha
See also: http://linux-ha.org/ReportingProblems

Reply via email to