CPU Load under Linux?
Hi Folks,
We're in the process of evaluating OpManager as a replacement to a hodge-podge BigBrother+MRTG system, and so far am pretty impressed (I'm assuming that the insane slowness of the system is due to the PIII-800 It's currently running on while we evaluate the software..). We're seeing a problem with the CPU Utilisation monitoring on our Linux (debian) servers however..
The servers are all running Net-SNMP 5.1.2, and OpManager is querying against 1.3.6.1.4.1.2021.11.9.0, which translates to ssCpuUser, which technically is only a small part of the system CPU utilisation, but I'd be happy just to have that part working for now... Querying that variable *always* returns the same value on a given server.. The MIB docs suggest that the variable is actually deprecated and we should instead be querying ssCpuRawUser, though I note that the value of that also never changes..
I'd actually prefer to be querying the 1-minute CPU Load average, which gives a much more useful value of CPU Load (e.g. on our backup cluster, the CPU load avg sits around 2.0 due to lots of network/disk traffic, but 'user' load would generally be 0).
I've tried adding a custom monitor for the appropriate path (.1.3.6.1.4.1.2021.10.1.3.1), however OpManager insists on adding a '.0' to the end when actually running the query.. Plus it doesn't seem to give me the nice speedometer-style display in the host snapshot.
Is there a way of either 1) modifying the default CPU Utilisation monitor to use my preferred OID (this would be preferrable), or 2) getting OpManager to not append .0 to the end of my custom OID ?
Thanks,
Damien
We're in the process of evaluating OpManager as a replacement to a hodge-podge BigBrother+MRTG system, and so far am pretty impressed (I'm assuming that the insane slowness of the system is due to the PIII-800 It's currently running on while we evaluate the software..). We're seeing a problem with the CPU Utilisation monitoring on our Linux (debian) servers however..
The servers are all running Net-SNMP 5.1.2, and OpManager is querying against 1.3.6.1.4.1.2021.11.9.0, which translates to ssCpuUser, which technically is only a small part of the system CPU utilisation, but I'd be happy just to have that part working for now... Querying that variable *always* returns the same value on a given server.. The MIB docs suggest that the variable is actually deprecated and we should instead be querying ssCpuRawUser, though I note that the value of that also never changes..
I'd actually prefer to be querying the 1-minute CPU Load average, which gives a much more useful value of CPU Load (e.g. on our backup cluster, the CPU load avg sits around 2.0 due to lots of network/disk traffic, but 'user' load would generally be 0).
I've tried adding a custom monitor for the appropriate path (.1.3.6.1.4.1.2021.10.1.3.1), however OpManager insists on adding a '.0' to the end when actually running the query.. Plus it doesn't seem to give me the nice speedometer-style display in the host snapshot.
Is there a way of either 1) modifying the default CPU Utilisation monitor to use my preferred OID (this would be preferrable), or 2) getting OpManager to not append .0 to the end of my custom OID ?
Thanks,
Damien