Forum Discussion

Kelemvor's avatar
Kelemvor
Icon for Expert rankExpert
2 months ago

Auto Balanced Collector Group vs Failover?

Hi,

I'm wondering if there are pros and cons vs setting up multiple collectors as an ABCG vs stand-alone collectors with failover.  If we have two collectors in the same location, monitoring the same servers, I'd use an ABCG.  However...

Let's say I have servers in Los Angeles and New York, and I have one collector in each location.  If I set them up as an ABCG, would LM be smart enough to have the LA machines be monitored by the LA Collector and NY by NY, or does it just assign them randomly?  I'm not sure if it has each collector ping each device to see which one has a better connection or anything like that.

I currently have the LA collector manually assigned to the LA machines.  I then have the LA Collector fail over to the NY Collector in the event of a problem.  Each location CAN see the other, but I don't want LM monitoring things across the connection when each location has it's own collector on-site.

Thoughts?

  • If you have collectors in different regions, you would want to avoid Auto Balanced. This would be more meant for traditional failover. As with traditional failover, your devices will be monitored by the collector you pick, generally the closest. But if something were to happen to that collector, the back up would try to take over.

    With Auto Balanced, while yes, the other collectors try to take over when 1 is down, you can end up with a collector with a horrible connection being the main collector. Auto Balance only works off of instance counts, and only balances when that threshold has been exceeded.

  • As far as I know, ABCG isn't that intelligent. nor does it mention along those lines at https://www.logicmonitor.com/support/collectors/collector-groups/auto-balanced-collector-groups. I wouldn't suggest doing ABCG across sites.

    Note, you may want to be careful about even normal fail-over to a different site. If a site goes down due to internet or power, all the devices would switch over the other site's collector. Since that collector is live but can't reach all those devices, it's going to alert on each and every device at that site as being down. I believe the Dependent Alert Mapping may help prevent that issue though.

  • If you have collectors in different regions, you would want to avoid Auto Balanced. This would be more meant for traditional failover. As with traditional failover, your devices will be monitored by the collector you pick, generally the closest. But if something were to happen to that collector, the back up would try to take over.

    With Auto Balanced, while yes, the other collectors try to take over when 1 is down, you can end up with a collector with a horrible connection being the main collector. Auto Balance only works off of instance counts, and only balances when that threshold has been exceeded.

    • Kelemvor's avatar
      Kelemvor
      Icon for Expert rankExpert

      OK.  That's what I thought, just didn't know for sure.  Thanks for the info.