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

I've recently bought a used rackmount 2950 server on Ebay to use for a specific project.

I'm going to be spending a lot of time on this server, so I want to try and identify any major problems before I have everything installed and running. That way, when it all goes to heck, and I have to buy a ton of new hardware, I at least know that I did what I could at the beginning.

So, my questions are:

  1. Are there any recommended tests to run (I'm already thinking fsck, but what else?) on the hardware before I install anything? This means harddrives, RAM, CPU, and whatever else.
  2. Are there any live disk utilities, or anything like that, that could automate the process?
  3. What should I monitor on an ongoing basis?

Thank you everyone for your time and attention!

share|improve this question
    
Use a configuration management system. Then you'll be able to get your stack set up in minutes, regardless of hardware or where it's running. –  EEAA Apr 22 '13 at 20:54
    
I don't think a configuration management system would be appropriate for a single server, unless it's faster to set up than the actual server. Do you have any specific systems that you are thinking of? –  MToecker Apr 22 '13 at 20:58
    
It is absolutely appropriate for a single server. It's not about being faster the first time, it's about having a known-good configuration you can apply very quickly to other systems, or to rebuild servers in the event of hardware failure. –  EEAA Apr 22 '13 at 21:02
    
Good advice, in general for a sustainable and economical datacenter. But, no plans at this time for deploying it to other systems, or anything similiar. I will look into it though, after going through the motions manually. Thanks! –  MToecker Apr 22 '13 at 21:12
    
@EEAA - that's assuming you a 2nd place to put your config management. Pretty useless if you have config management, but then only server where it's hosted (because you only have one) is the one that's been hosed... –  Mark Henderson Apr 22 '13 at 21:41
show 1 more comment

2 Answers

up vote 1 down vote accepted

Get your favorite battery of torture tests ready: memtext, badblocks, ... See if SMART has something to say. Check for any vendor diagnostic programs.

You should always keep an eye on the system's logs, and SMART.

share|improve this answer
    
Awesome advice! Thank you! What other tests would be in your toolbox? –  MToecker Apr 22 '13 at 20:59
    
On top of the above, a livecd with the "stress" tool is part of my checks of "pre-loved" server –  fukawi2 Apr 22 '13 at 23:48
    
@MToecker compiling a Linux kernel stresses quite a few things (CPU, memory, disk). –  Andrew Apr 23 '13 at 2:37
add comment

When redeploying older server hardware the major concerns are stability of PSU and motherboards. CPU and memory do not degrade as fast. The memory is ECC registered anyway. So no special memory testing is needed. Just check the system logs after a week of uptime. Storage should be redundant and it is considered consumables.

share|improve this answer
    
Ouch. Wasn't considering them to be consumables for this server, as it's not HA. I'm provisioning as RAID 5 for the database piece, and RAID 1 for the operating system. –  MToecker Apr 22 '13 at 21:11
    
what I meant was that passing whatever tests today does not guarantee that the same drive would not die tomorrow. So you should keep some spare disks in your inventory –  Alex P. Apr 22 '13 at 21:16
    
That I will do. At least one spare will be ordered this month. Thanks! –  MToecker Apr 22 '13 at 21:18
add comment

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.