HP servers HA solution

[kcpant]

Dear Friends,

One of my customer have 2 hp servers: ML 350 G2 with RH 8.0 & new ML 150 with SATA HDD. he uses ML 350 as mail server (sendmail). Now, his requirement is to install RH/ FC on new SATA based server and configure it as a HA member for ML 350 server, means, if 350 goes down, ML 150 should take it's place without any intervention and all mail data should be available.

Now, please suggest 1. which version of RH / FC can be installed on SATA based ML 150?
2. Can we make a HA solution for mail services on the servers having different OS versions?
3. Which components / configuration is required to accomplish this task ( I think I have to use some data sync mechanism which will replicate the data to backup (ML 150) server and some system which will trigger secondary system to work when main server is down)?

Please advise with some details as I am new to HA clustering thing.
Thanks in Advance..

[Steven E. Protter]

I think RH ES 3 or FC3 will work nicely.

You have two clustering options. Red Hat Clustering that ships wiht RH ES 3/4 or Serviceguard.

Both have advantages and disadvantages.

With Clustering Sendmail can be configured as a package that fails over from node to node providing HA for your users.

SEP

[kcpant]

Many Thanks for reply SEP,

Your answer cleared that I can install FC3 on this new server without any problem. Now, Please tell me will it work if I have different OS versions on both the servers? and can you guide me to some how-to links for clustering with some FC based GPL application.

Thanks again

[Ivan Ferreira · Accepted solution]

If you don't want to buy a clustering software, you can use heartbeat (www.linux-ha.org/heartbeat).

For a mail server, even when can be done with utilities like rsync, is not recomended to copy the files from server to server via network to sync the data. The data is changing very quickly. So you need a shared storage (Like MSA).

You must use the same linux version, upgrade you old RH 8 Linux.

You need two network adapters on each server, a serial port, and the shared storage, the same linux version and the heartbeat software.

[kcpant]

Thanks Ivan,

I appreciate the help porvided by you, any other such links or resources/ tips will be highly helpful for me to fully understand clustering. One more thing is, if I use rsync instead of a storage box, what complications will arise ( I assume the data on both nodes will not be exact, pl let me know if any other )

thanks again

[Vitaly Karasik_1]

These articles may be useful:

http://www.linux-mag.com/2003-11/availability_01.html Highly-Affordable High Availability
http://www.linux-mag.com/2003-11/drbd_01.html Data Redundancy By DRBD

http://h71000.www7.hp.com/openvms/journal/v5/#cluster - This paper surveys the cluster technologies for the operating systems available from many vendors

[Ivan Ferreira]

For Mail Servers, if you use rsync to sync the data, there can a problem depending of the MTA.

Sendmail uses one file (mailbox) for all messages for a user. The rsync utility is able to copy only modified files, then reducing bandwidht and sync time. But as sendmail use only one file, when a new mail arrives, then all the file must be copied again. So, rsync won't be too useful. Also, if the system crash while copying, the mailbox may be corrupt.

Other MTA's use one file per message for a user. This is recommended if you are going to use rsync for data syncrhonization.

[kcpant]

Closing threads open from a long time....