Ubuntu server creep - Virtualization? Cluster?

[LCPSWolf]

Hey all,

We started running 1 Ubuntu server for an open-source SIMS-like project for some special data collection - and we're now up to 3 with the addition of 2 mail gateways that provide email filtering.

Being the only Ubuntu guy here, I'm starting to wonder if I can do this better, as opposed to having 3 separate boxes currently.

Granted, I feel safer having 2 separate mail gateways, but expect I can get the redundancy in a different way.

That all being said - what's the best way to do this? Virtual servers? Cluster?

I've read through this post here:
http://www.edugeek.net/forums/networ...alization.html
And it makes it sound like virtual servers is the way to go. My only requirements would be: ease of use, redunancy / failover.

Thanks for any suggestions. I've only been running Ubuntu for about a year now, so I'm pretty new - I know it has some virtual pieces in it but haven't really looked at it yet (KVM?).

Thanks.
Damian

[CyberNerd]

The redhat cluster suite is in the Ubuntu repo's
redhat.com |

[CAM]

If you want redundancy, isn't it best to keep the two redundant systems on seperate boxes anyway? If something happens to the host box then both virtual systems go down.

[LCPSWolf]

Cam,

I tend to agree with you...paranoid, but sometimes that's for the best.

[teejay]

If you want redundancy, isn't it best to keep the two redundant systems on seperate boxes anyway? If something happens to the host box then both virtual systems go down.

Tbh, server hardware failure with decent servers is very rare these days.We had 15 physical servers and in the 7 years I've been here we had 2 failures of hard drives in RAID arrays which caused no downtime as they were hot swapped. What has caused nearly all our downtime is either environmental issues (air con failure, power failure etc) or software issues. Going virtual has reduced these problems dramatically by for instance pumping out a lot less heat, so our server room can now cope quite comfortably with an overnight air con failure and with greater UPS capacity due to less servers drawing load has a much longer uptime in case of power failure. Combine that with snapshots taken whenever any software is installed or updated so we have a working configuration to roll back to and we have a much more stable system.

[dhicks]

And it makes it sound like virtual servers is the way to go. My only requirements would be: ease of use, redunancy / failover.

I've found Xen on top of DRBD-mirrored storage volumes to be very reliable. It runs Windows Server as well as Linux-based VMs, although with Linux you have the extra trick of paravirtualisation availble, where the Linux VM runs a modified kernel that lets it know it is running as a VM and allows for better performance / less resource usage. I've found Debian to be rather easier to use that Ubuntu for both VM host and client, though.

--
David Hicks