Cheshire Cat Computing

Hi Steve,

Thanks for the tool! I'm having a bit of a problem with the CPU reporting. 4 ESX hosts are not reporting CPU usage ("Error reading CPU usage information.") and 1 host reports the same error sometimes and at other times has a CPU utilization of 103.09% or similar. 1 host is reporting CPU usage properly. This is with ESX 2.5.2 and check_esx version 1.4. Could you give me some tips to try and fix this?

I'm updating this in the next version.

Basically, if you have a twin CPU box, you can get up to 200% (IE, 100% of 2 CPUs). This is nicely confusing and misleading, which is why the next version will either correct this to a 100% maximum, or else give it in number of CPUs used.

The 'Error reading' can happen if no guests are running. This can result in part of the mib tree being unavailable, and it looks as if the ESX mib is missing, giving the error.

_________________
Steve Shipway
UNIX Systems, ITSS, University of Auckland, NZ
Woe unto them that rise up early in the morning... -- Isaiah 5:11

Thanks for the explanation.

After running check_esx for a couple days now all of the machines that were giving the error reading initially still do most of the time. All 4 have at least 12 VMs running on each. When host system load is light the the CPU usage sometimes is reported.

All of the host machines have 4 CPUs.

This sounds like a timeout problem in the reading. The new version (eventually) should give a better error message in this case. In the meantime, the $TIMEOUT variable in the code can be increased to something higher and see if this fixes the problem.

_________________
Steve Shipway
UNIX Systems, ITSS, University of Auckland, NZ
Woe unto them that rise up early in the morning... -- Isaiah 5:11

Thanks but I don't think it is a timeout issue. I went ahead and bumped up the value to 45 but when I run the command by hand it comes back almost immediately.

I don't mean to bother you but this is the last little bit left of my nagios rollout. As for the new version; will it check the disk space of vmfs volumes?

At themoment, I'm on holiday (until May) so I can't do any work on it.

The problem with getting this info is that I'm trying to NOT install any new software on the ESX server - IE, to do it all via SNMP. The ESX SNMP MIB is woefully limited. I've managed to get around most of this by having a special snmpd.conf you need to use (it uses special features inthe SNMP daemon to grab the contents of certain more helpful files under /proc) but in order to get such things as vmfs volumes its a bit harder.

Yes, it can be done, but it requires a whole new slew of stuff.

This is in the to-do list for after I get back to New Zealand and clear the 'real work' backlog.

I'm also working on another more helpful program (bundled in with routers.cgi) that dynamically sets up some dummy .cfg files and gathers the data - much better gaphs result, but its more complex. It turns out that getting *meaningful* ESX stats is surprisingly awkward.

_________________
Steve Shipway
UNIX Systems, ITSS, University of Auckland, NZ
Woe unto them that rise up early in the morning... -- Isaiah 5:11