Query On Redhat Cluster

[Md. Minhaz Khan]

Dear Experts,

Say, we have create a restricted failover domain for application "X" between Node A & Node B. We have set Node A as priority 1 & Node B as Priority 2.

Now consider the scenario, Node A become unresponsive (It becomes hang), Node B fence Node A. Now node B take over application "X".

My question is how do I make sure that application "X" do not switchover automatically to the Node A after reboot until we will manually fail over application "X" from Node B to Node A using command "clusvcadm".

Another Query, Is it possible to configure standby LAN card in Redhat Cluster like
MC/ServiceGuard ??

**Note: We will use RHEL 5.3 64 bit.

Thanks
Minhaz

[Matti_Kurkela · Accepted solution]

In common cluster terminology, this is called "preventing automatic failback".

If you're configuring the cluster using Conga, check the checkbox labeled "Do not fail back services in this domain." in the failover domain configuration. It does exactly what you're asking.

Please see my attachment in this thread:
http://forums.itrc.hp.com/service/forums/questionanswer.do?threadId=1452915

If you're editing the cluster configuration file directly, the option you need is 'nofailback="1"' in the failover domain configuration.

Please see:
http://sources.redhat.com/cluster/wiki/FailoverDomains

> Another Query, Is it possible to configure standby LAN card in Redhat Cluster like
MC/ServiceGuard ??

In Linux, this functionality is in the basic OS, not in the cluster suite or a separate package. It is called "bonding".

In principle, it is similar to APA on HP-UX. Like APA, it is more advanced than the standby LAN functionality offered by Serviceguard on HP-UX.

Please see:
http://www.mjmwired.net/kernel/Documentation/networking/bonding.txt

http://docs.redhat.com/docs/en-US/Red_Hat_Enterprise_Linux/5/html/Deployment_Guide/s2-networkscripts-interfaces-chan.html

The first link is generic Linux kernel documentation and describes the things you can do with the bonding feature, the second describes configuration procedures specific to RedHat.

For simple standby LAN functionality, you'll want mode 1: "active-backup".

MK

[Md. Minhaz Khan]

Dear MK,

My last query is, if i have done some changes in the "cluster.conf.new" file then run

#ccs_tool update /etc/cluster/cluster.conf.new

then it will update all members of a cluster. After running "ccs_tool" whether or not i need to restart cluster services (cman, rgmanager) etc...??

Thanks
Minhaz

[Matti_Kurkela]

> After running "ccs_tool" whether or not i need to restart cluster services (cman, rgmanager) etc...??

It depends on what you've changed. In general, most changes don't have a restart requirement.

Adding and deleting resources/services/failover domains will not require restarting any cluster services.

If you're adding a new cluster member, then you'll obviously have to start cman & other cluster services on the new member once it has been successfully added to the cluster.

Some specific cluster timeout parameters (quorum_dev_poll and totem token) will require a complete restart of cman on all nodes before their changes will become effective.

Please see:
https://access.redhat.com/kb/docs/DOC-2882
(RHN access required. The title of the article is about qdisk, but it contains important information on other cluster timeouts too.)

MK

[Md. Minhaz Khan]

Thanks MK, I have got my answer.

BR
Minhaz