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 poweredge 1855 blade server with 10 blades

I am trying to setup the system to handle:

  • 1 plesk server for mail and dns + dns replication to another machine
  • 1 system for httpd & mysql
  • 1 system for image hosting 300gb data

I was thinking to have

  • 1 plesk server 146gb
  • 2 hearbeat load balancers 73gb
  • 3 webservers with db replication 146GB
  • 3 image servers with rsync replication 300GB

Is there a better way to set this up would it be better to cluster all the drives and processers to work like one system somehow or should I keep the systems separate.

share|improve this question
    
Where does the 300Gb of data come from? I'm fairly sure the 1855's only had 147Gb disks... –  Mark Henderson Feb 5 '10 at 22:04

3 Answers 3

Without getting into the details, my $0.02 is you won't get high availability unless you buy another blade chassis because that is a single point of failure. And probably put it in a different rack.

share|improve this answer
2  
And put the rack in a different building, running off a different power and network grid and on a different tectonic plate. Depending on how seriously you take your HA attributes, of course. But HA is only mentioned in the title; in the body of the question the OP is only asking for load balancing. That could be done in the single blade chassis. –  mpez0 Feb 5 '10 at 21:35
    
@mpez0 I don't see anything about loadbalancing it just says "which is better" –  Jim B Feb 5 '10 at 23:58
    
@Jim B - He says 2 heartbeats –  MDMarra Feb 6 '10 at 0:33
    
The loadbalancing is in the servers, it's a question of whether the servers that are doing the loadbalancing are in the VM solution or standalone. –  Jim B Feb 6 '10 at 16:25

You could build them as a 10 member VMWare cluster, then run all your servers as VMs, this would help with availability a lot and allow for smooth scale-out as required.

share|improve this answer
    
This is exactly what we did with a Dell 1855 chassis. 10 VMWare hosts with DRS, plus an iSCSI SAN so that if one blade fails everything can be migrated to another blade and rebooted. Incredibly expensive though, the VMWare licensing was about 3x the cost of the fully populated chassis. –  Mark Henderson Feb 5 '10 at 22:02
    
And because the 1855 blades only have 2 NICs by default (unless you're a lucky one with another dual-nic backplane and all the daughter cards to make a total of 4 NICs) you can't use VMWare HA for hot avaliability either. –  Mark Henderson Feb 5 '10 at 22:03
    
The 1855's aren't certified for ESX4 so Fault tolerance\HT is a total non starter. –  Helvick Feb 5 '10 at 22:11
    
@Farseeker: You can use VMWare HA/DRS. You just can't configure it to best practices. –  Zypher Feb 5 '10 at 22:13
    
@farseeker, glad I stuck with HP blades, although couldn't you just use vlan tagging to sort the lack of NICs? –  Chopper3 Feb 5 '10 at 22:43

I think you might be stuck keeping things seperate. You could go vmware but your costs would skyrocket. You could probably buy an additional chassis for the cost of vmware (or just about any hypervisor) licensing. Other than hyper-v I'm not sure I'd trust any of the other virtualization platforms. It wouldn't hurt to look at KVM but from the literature it appears to only support redhat.

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.