[Openais] corosync starts too early

Vadym Chepkov vchepkov at gmail.com
Mon Jun 7 05:56:25 PDT 2010


On Jun 7, 2010, at 12:54 AM, Fabio M. Di Nitto wrote:

> On 6/7/2010 6:47 AM, Vadym Chepkov wrote:
>> 
>> On Jun 7, 2010, at 12:34 AM, Fabio M. Di Nitto wrote:
>>> 
>>> There is also a major catch-22 in changing them.
>>> 
>>> In certain environment it is necessary to start corosync right after the
>>> network because everything else in the boot process could require for
>>> example a cluster filesystem available (and corosync running as backend).
>> 
>> I realize that, that's why I brought it up.
>> But it would be certainly strange to see somebody starting resources relying on cluster's filesystem 
>> outside of cluster configuration.
> 
> Not really no. It´s actually common to mount gfs2 via fstab and start
> web servers outside a resource-manager controlled environment.
> 
>> And there is no sensible cluster configuration out there 
>> that will start file system right away,
>> some synchronization/DC detection has to occur and
>> by that time rc script has already finished.
> 
> hmm sorry, I really don´t understand this statement. Cman for example
> starts right after corosync (or starts corosync to be more precise) and
> at the end of the cman init process, cluster is fully available by the
> end of the init script.

My bad, I forgot corosync can be used for something beside pacemaker.
It would be nice to prevent this from happening though:

ls /etc/rc3.d/*corosync*
/etc/rc3.d/S99corosync

yum -y update
Updating:
 corosync                       x86_64                    1.2.2-1.1.el5                     clusterlabs                    148 k
 corosynclib                    x86_64                    1.2.2-1.1.el5                     clusterlabs                    170 k

# ls /etc/rc3.d/*corosync*
/etc/rc3.d/S20corosync

Is it corosync's rpm fault? Why priorities getting reset?

Thanks,
Vadym




More information about the Openais mailing list