On Thu, Apr 16, 2009 at 07:37, MAHESH, SIDDACHETTY M (SIDDACHETTY M)
<[email protected]> wrote:
> Hi List,
>
>
>   I am new to linux-ha and this is my first attempt at it.
>
> My configuration:
> 1. OS = Redhat Enterprise Linux 5.x
> 2. HA = v2.1.3-3 RPM install (using CentoS repository rpms)
>
>
> I am having trouble coming up with a clean cib.xml (see attachment).
>
> I want to implement a HA solution to satisfying the following
> conditions:
> ------------------------------------------------------------------------
> 1. I have two nodes - PrimaryNode and SecondaryNode - both have dual
> NICs (eth0 and eth1). 'eth1' is used for cross-over cable and heartbeat.
> 'eth0' is main application interface.
>
> 2. The application consists of multiple services (about 14 services). I
> have simplified it in the attached cib.xml to two services - service1
> and service2. The services are grouped into a resource 'ha_group' as
> they have to be started/stopped in a particular sequence.
>
> 3. The two services update variables service1-state and service2-state
> based on their running state. If successfully started, *-state is set to
> a value of 100. On stop/failure, the value is set to 0.

Why?

>
> 4. The two nodes use a virtual IP - 10.10.1.120.
>
>
> The HA requirements are:
> -----------------------
> 1. The nodes should be in active/active configuration. All the services
> need to be running on both nodes (to avoid startup delay). Only the node
> that is bound to the virtual IP will service end users.
>
> 2. On heartbeat failure, network connectivity loss, the other node
> should take over.
>
> 3. If one or more application services (service1 or service2) goes down,
> they should be restarted upto N times. Only if they fail to run after N
> restart attempts, the other node should bind to the virtual ip. Note
> that it would be good if the group is shutdown on the primary node - but
> not a hard requirement.
>
>
> Where am I stuck?
> ----------------
> 1. How to implement the "retry N times before migration" logic? I am not
> clear on this. Has anyone encountered this scenario? If so, is there a
> recommended way to approach this problem?

Upgrade to pacemaker 1.0.x and set migration-threshold=N
Seriously, prior to that we had the resource-failure-stickiness mess
that was completely unusable and I'm glad to see it gone.
_______________________________________________
Linux-HA mailing list
[email protected]
http://lists.linux-ha.org/mailman/listinfo/linux-ha
See also: http://linux-ha.org/ReportingProblems

Reply via email to