[Linux-HA] HB2 with DRBD (master_slave) and OCFS2 (clone)

Andrew Beekhof beekhof at gmail.com
Mon Sep 17 03:14:17 MDT 2007


On 8/31/07, Raoul Bhatia [IPAX] <r.bhatia at ipax.at> wrote:
> hi,
>
> i have a problem with hb2, drbd and ocfs2 and two (master-master) nodes:
>
>
> 1) when i start one node:
> heartbeat is starting and activates the 2 drbd devices. most of the time
> they become master (primary) but sometimes even this fails (see attached
> config/logfiles)

i'll look and get back to you

> ocfs2 should wait until the corresponding drbd device is in master
> state and then be mounted but somehow (as far as i can see in the logs)

you need to specify that with a rsc_order constraint

btw.
     <rsc_order id="order_apache2" from="clone_ocfs2_www"
type="before" to="apache2_clone"/>

looks to be the wrong way around.

I recently added this to the DTD...
Read as:
     to_action to type action from

Which in your case is:
     start apache2_clone before start clone_ocfs2_www


> ocfs2 is started immediately after the drbd device and is not waiting
> for the drbd device to become master hence failing to start.

as above, because you didn't tell us to do so

> then the ocfs2 resource is never started again.

because it failed to start on both nodes
start failures are terminal

> manually activating/mounting the drbd and ocfs resources is no problem.
> i can even set the env variables and use the scripts to do that.
>
>
> 2) when i start both nodes:
> when i start both nodes, most of the time the drbd devices start
> successfully and are placed into primary/primary mode.
>
> ocfs2 fails to start on one node, failing over (or it is started on the
> second node) and is successfully mounted there.

again, this is due to the ordering issue

> so my question is: is there a configuration error in my cib.xml and/or
> is this a hb2 bug?

config



More information about the Linux-HA mailing list