Clustering guidelines for RHEL6

[rhel6_cluster]

Hi all,

I am a beginner, i am configuring clustering for the first time. Therefore could you please provide me some guidelines and pre-requisites for cluster setup in RHEL6 ?
I have 2 BL460 G7 blades, storage HP P6000 for quorum and shared disks and planning to use iLO for fencing.

Appreciate your help!
Thanks!!

[Matti_Kurkela · Accepted solution]

Sorry Senthil, looks like I missed your last post in this thread. Anyway...

1.) apparently you can get the information with "mkqdisk -d -L"

2+3+4.) deadnode_timeout is a RHEL4.x Cluster parameter, and does not exist in the same form in RHEL5.x and newer.

The closest equivalent is the combination of qdisk's interval and tko parameters: the qdisk cycles once every interval seconds, and once a node has failed to participate or its qdisk heuristics score has been too low (it must score more than 50% of the sum of its heuristics scores) in tko consecutive cycles,  the node will be declared dead.

When a node is declared dead, an eviction notice for it is written to the shared quorum disk, and fencing is triggered. Note: any failover actions can only happen after fencing is successful.

In other words, the dead node will be timed out by qdiskd in (interval * tko) seconds. By adjusting these two parameters, you will adjust qdiskd's node timeout.

---------------------------

rhel6_cluster:

It would be better if you started your own thread instead of resurrecting an old one for your question. Click on the blue "New Post" button at the top of the topic list for the Linux -> System Administration section.

A few points:

• Use bonding to make your network connections fault-tolerant.
  
• IP Multicast must work correctly or else RHEL Cluster won't work. This is an absolute requirement.
  
• Use RHEL 6.1 or later, not 6.0. If you must use RHEL 6.0 installation media, register the systems to RedHat and update them to get to the latest update levels before starting to configure the cluster. 
• A two-node cluster without a quorum disk may get stuck in a reboot-and-fence loop, where each node gets fenced, reboots and then fences the other node if the network connection to the server OS has failed.
  

[rhel6_cluster]

Thanks!

I much sizing is required for quorum disks? also how to add shared disks in the cluster, so that both the nodes will read it? just mounting disks or any special configuration to be done?

May i know how to configure iLO fencing in RHEL 6.1/2/3 ??

Thanks!
Much appreciated.