[Linux-HA] Resources launched on both nodes, and unwanted "unset DC"

[Erwan Le Gall]

Erwan Le Gall a écrit :
> Hi,
>
> I'm trying to setup a two nodes failover cluster on an up to date Debian 
> Lenny with Heartbeat "2.1.3-6lenny4". There is various resources to 
> handle, but I need to have all of them on the same node at a time and I 
> need them to start in a specific order.
> I'm currently using a lot of LSB scripts since I'm unable to get the ocf 
> ones working for the moment (but that is maybe all my problem).
>
> First problem: I launch heartbeat on the first node, but I have the 
> tomcat services which start before the other (I guess that's cause 
> tomcat LSB init script return a right status before be sure that tomcat 
> is really started).
>
> Except this problem, everything works well on the first node (even if 
> the tomcat process are marked "starded" a little too soon).
>
> Second problem, and the main one: when I launch heartbeat on the second 
> node, there is a strange behavior.
> The second node start some services in the same time than first node 
> (like apache for example). Then node1 stop some services, so node2 
> takeover, both nodes stop all resources and node2 become DC.
> The "both launched" servcies are apparently "tomcat" "Minig-Backend" 
> (which is a tomcat application) "Tina_Atempo" (which is a saves 
> program). In logs, I obtain "ERROR: See 
> http://linux-ha.org/v2/faq/resource_too_active for more information.".
> Weird thing, if I do a heartbeat restart on the node1, the duplicated 
> services stop_start but then node2 keep resources.
>
> I'm asking myself: why heartbeat try to launch the resource on passive 
> node? Is that's because the LSB scripts don't return a good status result?
>
> Note: I re-verify, this services are not launched before heartbeat start.
>
> Sorry if it's a common mistake but I haven't found the answer.
>
> There is my cib.xml : http://pastebin.com/f1fdf0988
>
>   
Ok, Like usually, I post and found during the post or almost. It's the 
"non LSB compliant" scripts problem.

Sorry for the noise.

-- 
Erwan Le Gall