7

I'm about to spec out some VMs to run Nagios out of. We will be monitoring 2000+ devices and upwards of 4+ services on each, including AD replication, LDAP, HTTP, SSH, etc.

If I were to create a somewhat generic post asking for ideas on how to spec the VM out (CPU, Mem, Disk, etc) based on scale, would that be acceptable? I'd also like to ask for people to point me to any resources as I've come up empty, other than finding minimum requirements, nothing to really help calculate when scaling.

  • 2
    You could always pop into the serverfault chat room and ask if someone there has experience in large-scale nagios deployments. We're much open to broad questions in chat (it's why the chat rooms were introduced). – pauska Jun 12 '14 at 19:24
  • Great point, @pauska Thanks for the idea and reminder. – TryTryAgain Jun 12 '14 at 19:32
  • Glad I came across this Q. I didn't realize there were SF chat rooms! – SturdyErde Jun 13 '14 at 14:28
10

Your question will almost certainly be closed as a duplicate of Can you help me with my capacity planning?.

This sort of question, and variants of it, pop up so frequently that the community generally does not want to post the same answers over and over, thus the three canonical questions about capacity planning. Almost all such questions are closed as duplicates of one of these three questions. (The other two focus specifically on web servers and databases.)

Since the answers to these questions explain how to determine what the correct scale is for virtually any particular use case, it's unlikely that this question would last very long.

  • 2
    I appreciate it. I think my best approach may be testing with a smaller set and seeing how resources are taken up by specific monitors. I was sure I'd get an answer like this. Unfortunately, the existing questions/answers didn't help me all that much (I'm sure everyone writing duplicates feels that way). Thanks for confirming. – TryTryAgain Jun 9 '14 at 22:43
  • 1
    Yes, start small, and watch utilization as you add in chunks. Few systems scale linearly, so don't be surprised if you hit a wall. Monitoring systems are particularly because it is hard to make them completely orthogonal to the systems that they're monitoring. (For instance, adding a lot of checks to lots of systems could begin flooding your network layer.) – gWaldo Jun 13 '14 at 12:45

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .