Major graphs in gaps and timeouts with SNMP Informant

[stormonts]

We had been monitoring our Windows servers using WMI, but didn't like how that method was filling up the logs on the Windows servers with all of the various logons, logoffs, etc.

Recently we installed SNMP Informant Free on a number of Windows servers and now I'm seeing an incredible amount of gaps in the Windows graphs. We currently have Cacti .8.7e, 121 hosts, 3,484 SNMP items, are using Spine (using 30 threads), have a 1 minute polling time, and our average poller runtime is 37 seconds.

A number of servers have started giving "WARNING: SNMP timeout detected " errors and the occasional "WARNING: Poller Output Table not Empty" message. The devices in question originally had SNMP timeouts of 500 and have been raised as high as 7,000 and they get 15 maximum OIDs per request. However, running a SNMPGET for the OID in question returns an answer within 1 second, so I'm at a loss as to why Cacti is taking so long but the command line gets the result immediately.

[stormonts]

Just upgraded Cacti to .8.7g and also upgraded Spine to .8.7g and am having the same issues. It only happens with some of the Windows servers, but will happen with them consistently.