0

I have a couple of old Dell PowerEdge 2900 servers that have come free and I would like to use them to host a mail server cluster.

They are a bit old, but have been running workhorses for a while, and we have a 3rd one that we can use for spare parts. In fact, I took drives out of the 3rd one and split them between the 2 servers that I am going to use.

I've configured two of them with 8GB of memory each and 7 x 146GB SAS 15K drives. I've created that as a single RAID5 array. So total disks in each one is 7.

I've installed VMware ESXi 5.5 on them and setup an Ubuntu Server and a couple of Windows 2012 VM guests. In VMware, the entire RAID5 array is configured as a single datastore.

The plan is to use them as a cluster with Ubuntu + Zimbra and then the 2 x Windows Server VM guests are going to be Veeam proxys.

Everything works, but the VM guests seem unreasonably slow.

I've double checked that the array controller batteries are both in working order, and I don't have any other Dell hardware warnings to indicate any kind of problem.

Questions:

  • Has anyone else had any performance problems on similar hardware?
  • What are the most number of disks you should put into a RAID5 array for a VMware datastore?
  • Is there a more efficient way to configure the disks in terms of array volume type and size, and/or datastore size?
  • Are there any other things, besides the array battery, to check that could be hampering efficiency?
David Killingsworth
  • 101
  • 2
  • 12
  • 6
    You mean you put load on slow servers and now you complain the servers are slow? Am I missing something obvious here? Voting to close- this is for professional system administration questions only. – TomTom Nov 12 '14 at 12:07
  • 1
    The typical *guest* in our ESXi systems has 8 GB of RAM. For that matter, so does our typical workstation. And to answer your question about the most number of disks you should put in a RAID 5 array, the answer is 0. – HopelessN00b Nov 12 '14 at 13:46
  • `Everything works, but the VM guests seem unreasonably slow` - Please elaborate. How are they slow? What performance diagnostics/benchmarking have you done? – joeqwerty Nov 12 '14 at 15:52
  • Thanks for the encouraging comments. I should have indicated that they have Intel Xeon processors in them. Their roles previously were stand alone MS Server 2003 with SQL server 2008 servicing about 80 workstations for a medical records software application. Those servers were virutalized on other VMware hosts, which freed these guys up. In an SMB/SME environment, we have ran ESXi servers with about the same resources and had no problem. By slow, I mean that the VM guests are very sluggish, to the point of unusable. I suspect it's disk access that is the bottleneck. – David Killingsworth Nov 12 '14 at 15:55
  • joeqwerty, part of the purpose for posing this question here was to get ideas on what I should be checking, method for performing diagnostics and benchmarking. Do you have any recommendations so that I can provide more scientific data? – David Killingsworth Nov 12 '14 at 16:00
  • How do you know it's a storage problem? Maybe it's CPU or memory over-provisioning or something else. Google for esxtop and you'll find tons of information that'll help you. – Mario Lenz Nov 12 '14 at 18:19

1 Answers1

1

You're right on the edge of what's acceptable for an ESXi 5.5 host. The processors on that host are about 5 generations old. RAID 5 is also a bit of an issue, but if you have an array battery and write caching enabled, it shouldn't be terrible.

I'd say that the gear is too old to be used for a new deployment. 8GB of RAM is tight. How have you apportioned it to the guests?

More than anything, how are you defining "unreasonably slow"? What processors are in use?

Finally, there could be a slight chance that you have a VMware revision issue. What build number of ESXi are you using, and are you willing to update to the most recent?

Community
  • 1
ewwhite
  • 194,921
  • 91
  • 434
  • 799
  • I should have been more specific. I have 5.5 Update 1 installed. I could update to the latest or re-install an older 5.1 esxi if that would help. I really suspect that's is the disk array that is slowing down everything. – David Killingsworth Nov 12 '14 at 15:47
  • There are vmware patches between releases. If you didn't apply them, you're definitely out of date. Again, can you confirm that you have write cache on your array controller? – ewwhite Nov 12 '14 at 16:02