trapni at gmail
Jun 11, 2012, 9:28 AM
Post #1 of 1
(137 views)
Permalink
|
|
A highly available central nova-network
|
|
Hi all,
while I am very happy with our OpenStack cluster, I am now about to fill
the last gaps with good mood,
that is, implementing high availability to the central cloud controller, as
well as the central nova-network node.
The latter is now of my interest, and I was thinking forth and back,
whether or not to use
keepalived, heartbeat, or corosync+pacemaker.
Again, I am much in favor of the latter, however, but I am not sure whether
the choice I made,
is truly the right, and I am asking here for opinions, primarily targeting
at you, who have also been
deploying OpenStack at (kind of) large scale with a central but highly
available configured nova-network node.
I am having 2 dedicated network hosts, currently only one active, and I'd
like to take the second
as hot backup, but wonder how well OpenStack plays with what I have in my
mind.
In fact, `nova-manage service list` shows the following:
root [at] colossus0:~# nova-manage service list | grep nova-network
nova-network cesar1 nova
enabled :-) 2012-06-11 16:24:38
nova-network cesar2 nova
enabled :-) 2012-06-11 16:24:42
the first one is the primary active, the second is the secondary node, that
ought to take over when the first goes down.
I may not run nova-network on both nodes at the same time, am I right ?
For example because it takes over floating IP addresses - are there any
other reasons why
I am now allowed to run nova-network process on both nodes at the same time?
OTOH, what exact HA solutions are *you* using? Is anyone of you using
Pacemaker/Corosync, and if,
how did you configure it? Or what else did you use ?
Many thanks in advance,
Christian Parpart.
|
