Dejan Muhamedagic wrote:

>>> When one PostgreSQL server fails, the setup will still work fine. When 
>>> the failed PostgreSQL instance is back, the data should be first 
>>> "synchronized" from the running PostgreSQL server to a server which was 
>>> failed a while ago.
>>>
>>> It is best if such a script could be started by Heartbeat running on the 
>>> active node, as soon as it detects that the other node is back.
>> If you need such thing - I'd personally be most comfortable with not
>> starting the cluster at boot time. Then you can do whatever you need to
>> do and then - when you _know_ everything is right, the script is done
>> etc. - start the cluster software.
>>
>> Just my personal preference.
> 
> It would be mine too. Nothing wrong to have a script for that,
> but best to run it by hand so that you can check the database.
> Besides, your nodes shouldn't be disappearing that often.

OK, then I would like to be at least notified that heartbeat lost 
connection with the peer, and that it re-established the connection again.

Can I at least have that with heartbeat?

Monitoring (i.e. nagios) will usually only check the server every 5 
minutes or so, so a "spontaneous" reboot (or restart on kernel panic) 
may not be detected by that (yes, I can send a mail via startup scripts, 
but I'd rather heartbeat told me about its problems).


-- 
Tomasz Chmielewski
http://wpkg.org
_______________________________________________
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