[Openais] corosync starts too early

[Fabio M. Di Nitto]

Hi,

On 6/7/2010 6:27 AM, Vadym Chepkov wrote:
> Hi,
> 
> I think corosync starts too early during system initialization.
> Current priorities in init.d script seems to be wrong:
> 
> corosync-1.2.2-1.1.el5:
> # chkconfig: - 20 20
> 
> I observe very strange behavior, if it starts as configured I get this error
> 
> # corosync-cfgtool -s
> Printing ring status.
> Local node ID 319425034
> Could not get the ring status, the error is: 6
> 
> But if I restart it, all is well:
> 
> # corosync-cfgtool -s
> Printing ring status.
> Local node ID 352979466
> RING ID 0
> 	id	= 10.10.10.21
> 	status	= ring 0 active with no faults
> RING ID 1
> 	id	= 10.10.3.21
> 	status	= ring 1 active with no faults
> 
> I think the start should be pushed way down, after ntp starts (58) , for sure
> heartbeat's priority is more sensible:
> # chkconfig: - 75 05
> 
> but I would push it even further, after sendmail (80), possibly
> I personally do sed -i -e 's/.*chkconfig:.*/# chkconfig: 345 99 00/' /etc/rc.d/init.d/corosync in my kickstart :)

to keep this a short discussion, the current values were never tested on
RHEL-5.

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).

It´s possible that we might have to ship a specific rhel5 init script
with different values as solution to this problem. Changing the current
generic init is not an option as it would affect too many distributions
vs a specific one.

Fabio