Take the 2-minute tour ×
Server Fault is a question and answer site for system and network administrators. It's 100% free, no registration required.

I have a task where I need to build a failover cluster in two cases: first with servers on Red Hat Enterprise 5.1 and second with SUSE Linux Enterprise 11 SP1. Both cases have SAN. I know there are many ways to build failover cluster, but I can’t find out more, so I need next:

  • The ways to build it? I know only virtualization.
  • Any good book or resource to broad my mind?
  • I’ll be glad to hear any suggestion.

Thanks!

EDIT #1: failover of servers with bussiness application on it.

EDIT #2: will be great to hear summary about solutions with SLES servers?

EDIT #3: So if I understand correctly, in my cases the main ways are to use internal solutions or virtualization. So now I have additional questions:

  1. Does manufacturer of blades provide some solution? For example HP or IBM.
  2. (Without virtualization) Do I need additional server to control "heartbeat" between main and redundant servers?
  3. (Virtualization) For example I have several physical servers with VMs. Do I need additional server to control availability of VMs and to move VMs to another physical server in the case their physical server failure?

Sorry for my poor English.

EDIT #4: Failover of VM or OS on physical server. In both cases will be used SAN , it's not specified, but I think with file system image on it. I started to think that my question is stupid and I need to remake it.

share|improve this question
1  
Failover of what? The answers are somewhat different if you're failing over Virtual Machines or something like Apache/MySQL. –  sysadmin1138 Jun 19 '12 at 12:07
    
Small tip: failing over virtual machines is by far easier than services on one host. I use xen VM hosts with DRBD devices as backends for the VMs, which works well. –  Halfgaar Jun 19 '12 at 12:59
    
Failover? Except for very some very niche applications, implementing high availability via failover is such a bad idea and so last millenium. Load balancing concurrent services makes a LOT more sense. –  symcbean Jun 19 '12 at 14:09
    
RH 5.8 is current. So is SLES 11 SP2. Is your task to failover virtual servers or is it enough to fail over the application from one physical machine to the other? –  Nils Jun 19 '12 at 20:27
    
Please provide more details about the application. –  Nils Jun 19 '12 at 20:28

4 Answers 4

up vote 0 down vote accepted

Clustering, virtualization and HA are "internal" solutions for RedHat 5:

  • XEN
  • Heartbeat

are part of the distribution.

With SLES11 HA is an "external" solution, since the HAE is a license of its own. But: SLES11 already contains XEN and KVM - you are free to choose.

Since you have SAN LVM is possibly all you need (else DRBD is part of RH5 and SLES11-HAE).

I dislike SLES11-HAE since it does not contain the full "old" Heartbeat-Stack any more, but sports a multicast based HA-solution.

Instead of buying the HAE license you can download and compile heartbeat yourself...

share|improve this answer
    
Do I need additional server to realize High Availability? –  light Jun 19 '12 at 20:42
    
@light HA begins with two physical servers. You can simulate a cluster with two virtual ones - but if they are located on the same physical server this is not HA. –  Nils Jun 19 '12 at 21:06
    
In my cases I have two physical servers, main and redundant. So, I don't need third server to control availability of main server using solutions you described in your answer. Do I right understand you? –  light Jun 19 '12 at 21:16
    
@light two servers are enough to start with HA (and the mentioned software). There are clusters out there that require an additional quorum server - that server is ALWAYS on the wrong side (Murphy`s law). –  Nils Jun 20 '12 at 18:43

To be direct, documentation is your friend... http://docs.redhat.com/docs/en-US/Red_Hat_Enterprise_Linux/6/html/High_Availability_Add-On_Overview/index.html

In addition there is the RH536, Red Hat Enterprise Clustering and Storage Management class: https://www.redhat.com/training/courses/rh436/

Clustering, HA and load balancing is not a simple topic and will take some effort to truly appreciate. Every application is different when it comes to clustering as well. For instance not every application can run two instances at once, meaning each instance thinks that it has exclusive access to the database. This would be a candidate for Active/Passive or Hot/Cold HA clustering.

At some point though you will need to just dive into this and start experimenting. Be sure to keep notes so you can more easily build your production test system when you're ready.

share|improve this answer

If You want tools that will work with different linux distributions there are several options (depending on what You need):

  • Heartbeat Linux-HA - link
  • HAProxy - link1 link2
  • Red Hat HA - link
  • SUSE Linux Enterprise High Availability Extension - link

As for virtualization Vmware HA Cluster is the way to go.

share|improve this answer

I had to do both. I built two node clusters on both SLES 11 SP2 and Redhat 6.4 (Centos 6.4). Shared IP, storage, and application (Oracle). SLES is much easier. Still takes awhile, but much easier. Redhat 6.4 is a lot more tedious and took quite a bit more skill.

Overall, Redhat was much more complicated to get up and running. A lot more fine tuning, especially with quorum disk. In fact it is quite tedious to the point where when I googled around just about every answer I saw said to forget the quorum disk and go for a split-brain race (not ideal).

I finally did figure out how to get the quorum disk up and running, very handy because you can add additional metrics/tests to it easily (like ping the default gateway)

We did not do any performance tests of SLES vs Redhat, so I can't tell you which one is faster. I can't imagine they are far off from each other. Redhat definitely was harder to build. However, considering CentOS = Redhat and CentOS = free, if you want a free enterprise level cluster solution, it's good to know how to do CentOS/Redhat (a.k.a. Oracle Linux also)

I just put up a brief tutorial plus detailed ebook on SLES 11 SP2 clustering here: http://geekswing.com/geek/building-a-two-node-sles11-sp2-linux-cluster-on-vmware/

I have something similar for CentOS/Redhat, just need to polish it up.

Hope that helps!

share|improve this answer

Your Answer

 
discard

By posting your answer, you agree to the privacy policy and terms of service.

Not the answer you're looking for? Browse other questions tagged or ask your own question.