Virtual Infrastructure Monitoring Accurate?

The accuracy of the Orion Virtual Infrastructure Monitoring has recently come into question. Recently our entire Orion infrastructure was restarted (database, main poller, additional poller, and additional web) and after being restarted the perfomance data for all of our clusters is significantly different. Below I have attached CPU load for two different clusters to show what I am talking about.

Other than the restart of Orion, nothing else changed at that time. This has lead me to believe that the Virtual Infrastructure Monitoring numbers may not be accurate.

I would am really hoping that somebody can shed some light on this for me.

Find more posts tagged with

Accepted answers

byrona

In case anybody is still following this...

In a later post located HERE I did determine that the cluster level data being polled by IVIM is not (or at least not always) correct. The development team is going to be updating the polling method as I indicated in that thread.

All comments

byrona

Does anybody have any idea what might be causing this behavior?

fcaron

Never seen that.

I'll let other chime in but I would recommend opening a support case, Byrona.

I will also ask around and post here if I find something.

Something you can try, is polling these OIDs against these devices, with an SNMP tool and see what the device responds with, now (i.e. after the reboot). Won't tell you the whole story, but this wil be a non-Orion data point.

netlogix

The node, is it the ESX/ESXi server? Or a guest?

Have you looked at any of the charts in vCenter?

If it was an ESX/i node, then maybe some heavy server v-Mo'd off that Node to a different one. Did you recently apply any ESX/i patches?

byrona

It's actually a Cluster.

byrona

Never seen that.
I'll let other chime in but I would recommend opening a support case, Byrona.
I will also ask around and post here if I find something.
Something you can try, is polling these OIDs against these devices, with an SNMP tool and see what the device responds with, now (i.e. after the reboot). Won't tell you the whole story, but this wil be a non-Orion data point.

I am avoiding opening a support case on this because (from my experience) this is the type of issue support won't really know what to do with. I will likely end up uninstalling and reinstalling a bunch of stuff to no end. The problem is there isn't anything specifically wrong (errors, etc) for them to focus on and for that same reason we won't know for sure if/when it's fixed.

I just really have to question the accuracy of the data from Orion on this when there is such a sudden change when nothing changed on the cluster, only Orion getting rebooted which should not have this type of impact on the cluster. To me it all strongly points to a data accuracy issue.

byrona

Something you can try, is polling these OIDs against these devices, with an SNMP tool and see what the device responds with, now (i.e. after the reboot). Won't tell you the whole story, but this wil be a non-Orion data point.

I am pretty sure Cluster utilization data is pulled through the VMWare API and not through SNMP so this won't work.

netlogix

Oh, % use on that can be effected by maintenance mode on the esx hosts, or upgrades of esx hosts, or new esx hosts added to cluster.

I recently upgraded two of my hosts from PE2950 to R710's and my graph looks a bit like yours.

But when you look at the graph in vCenter, you can see that the CPU MHz stayed about the same, but the total pool increased which lowered the percentage. Maybe there is a resource that you can show your customer that is the MHz used not the percentage. (but of course your situation could be different from mine)

Karlo.Zatylny

You are correct that this is pulled through the API, Byron. Does the data look accurate now as compared with VIC? Or does VIC look like it was prior to the restart?

Thanks

byrona

In case anybody is still following this...