DX NetOps

Expand all | Collapse all

Alarm correlation for VMs to Hosts

  • 1.  Alarm correlation for VMs to Hosts

    Posted 10-19-2017 02:25 PM

    Is there anyone out there that is successfully monitoring the relationship from hosts (e.g. ESX) to VMs dynamically without using UIM since Spectrum discovery has no knowledge of the relationship of the host to VM?  If so, how do you do it?

     

    Here is our situation:

     

    If there is an outage/reboot 'upstream' from the ESX host (such as a distribution switch, FW, etc.) and the VMs are not fully mapped to the ESX host, all of the VMs will alert (ESX host will be suppressed if it is mapped to the switch) and create possibly 1000s of tickets ("DEVICE NOT RESPONDING TO POLLS"), depending on how many VMs are 'downstream.'  This causes AlarmNotifier to get backed up, Remedy gets backed up, BPPM gets backed up, and possibly XMatters gets backed up.

     

    Yes, we could manually map the VMs to the ESXs, but if VMotion or something equivalent happens, then the mapped environment is no longer correct.

     

    Since we use Alarm Notifier to pass the alerts to BMC's BPPM-->Remedy, we can't use the Condition Correlation tool since the alerts would be passed before the correlation kicks in (i.e. Alarm Notifier would do the 'set' of the 1000s of alerts before correlating into, say, one alert indicating a switch went down).

     

    Anyone doing anything creative in this area?

     

    Thanks.



  • 2.  Re: Alarm correlation for VMs to Hosts

    Posted 10-19-2017 02:35 PM

    The two GA methods for providing guest/host relationship information to Spectrum are via UIM with the vmware probe, as you mention and via Virtual Assurance for Infrastructure Managers (VAIM) with vcAIM.  I suppose you could try to get those relationships via vCenter APIs and then update the relationships when vCenter publishes an event that vMotion, new guest, etc. occurs since that's what the GA methods do.