Linux Clustering

[KarloChacon]

hi guys

I need to create a 2 node cluster Active/Passive

I was reading a lot about different stuff

DRBD = when there is no shared storage

Heartbeat = Cluster messaging layer

OpenAIS = Cluster messaging layer

Corosync = Cluster messaging layer

so I was thinking heartbeat was the way to go but it seems like it needs pacemaker in order to work since pacemaker will check the services to detect if the service if down and do a failover...

ok but I was reading an article about heartneat and saw this comment 

Oliver wrote:
You’re seriously using Heartbeat in R1 mode? Time to step a few years forward into the present and check out Pacemaker and Corosync.

source: http://www.mysqlfanboy.com/2010/07/mysql-ha-with-drdb-and-heartbeat-on-centos-5-5/

so my confusion is which cluster message layer do you recommend me?

heartbeat - openais or corosync?

thanks a lot

[gstanden · Accepted solution]

Hi,

I just finished doing alot of work with this experimenting with using it with various Oracle products.  First, here is a great link for doing a step-by-step project to get familiar:

http://www.markround.com/archives/archives/44-Building-a-redundant-iSCSI-and-NFS-cluster-with-Debian-Part-1.html

I found that the heartbeat cluster worked great based on working through this example above.  I was able to setup a Linux HA cluster across 2 linux machines which, when heartbeat was stopped, would failover the TGT san, the virtual IP's, and would switch the DRBD active to the surviving node.

Another link I found very useful was this one:

http://www.debian-administration.org/articles/578

In general the heartbeat clusters are much easier to setup and manage and work great.  If you go to pacemaker you have to use xml pages to configure them, and it's non-trivial, but if you like the challenge and that sort of thing, pacemaker is the recommended approach now. 

HTH

Gil 

[user_smanas]

following is the snippet from the Linux-HA Mailing List:

<snip>

Difference b/w Corosync and Hearbeat

"heartbeat" was used for many different things, among them
* messaging layer (= cluster communication)
* resource agents (= resource management)
* policy engine (= which resource is started where)

since a couple of years, these specific tasks have been split
into different (sub-)projects (e.g. pacemaker), and heartbeat has been
"reduced" to be one messaging layer. corosync (formerly also known as
openais) is another one.

nowadays, most people seem to prefer corosync + pacemaker over
heartbeat "v1" (legacy) or heartbeat + pacemaker.

The OpenAIS software is built to operate as plugins to the Corosync Cluster Engine

Pacemaker is a cluster resource manager. It achieves maximum availability for your cluster services (aka. resources) by detecting and recovering from node and resource-level failures by making use of the messaging and membership capabilities provided by your preferred cluster infrastructure (either Corosync or Heartbeat).

</snip>

Hope this helps

Regards

Maaz