Overlapping rules/monitors best practice

[anthmilic]

Hi all,

We have deployed the Veeam MP for SCOM 2007 R2 and I'm charged with tuning it. Is there a best practice for overlapping rules/monitors with the SCOM agent/Veeam when it comes to monitoring the basics of a VM like cpu, memory, disk and network? I'd like to disable unneeded rules/monitors in either the base OS MP from microsoft or the Veeam MP but not sure what's the best approach when I take into account things like reporting etc. We also definitely need the SCOM agent installed on all VMs to capture app data.

Thanks for any assistance.

[agolubnichy]

Hi Anthony,

The OS MP would do the monitoring from the perspective of Windows (or, in a genralized sense, from the OS perspective), while nworks MP does the monitoring from the perspective of the hypervisor. Thus, there will be no truly overlapping metrics: the OS data and hypervisor data are different pieces of information. So, the data coming from the agents inside the VMs will be processed by the OS MP, while the VMware-related data (coming from collector+agent machines) will be processed by nworks MP.

Please, let me know if this helps.

[sergey.g]

Hello Anthony,

Thank you for such a good question.

But I'm afraid there is no simple answer for it. Veeam MP provides you with data from Virtualization level of your infrastructure, which could mean for instance that VM is consuming 50% of allocated CPU, while inside VM, native MP could report about 100% cpu usage. This could happen if ESX cannot provide enough CPU resources to the VM (because other VMs are fighting for the same CPU, or VM is waiting for IO from datastore or some other operation is happening on the esx or storage). So sometimes it could be necessary to look inside the VM. Another Example: when you see that some VM creates huge disk IOPS, you have a choise: either move this VM to another dedicated storage device or to investigate what is happening inside the VM and if disk throughput of certain application should be investigated. For Memory it could be also very useful to have both, i.e. it could be highly dangerous for performance if your VM is experiencing memory swapping on ESX and inside the VM at the same time, the only way to track this is to have both monitors/counters enabled, inside the VM and in Veeam MP.

So, I think I would recommend you to have both enabled for a while, then you will discover if for certain VMs some internal counters can be disabled.

The only counter which is probably overlapping the SCOM agent inside the VM, is free % of virtual disk space. So, theoretically there is no need for such counter inside the VM.

Hope this helps.
Thanks.