SNMP Traps not sent during state changes

[tajensz]

Hi,

I'm having an intermittent problem with SNMP traps from vCenter and was wondering if anyone else has experienced something similar.

Certain alarms are configured to send a trap during state change (green-to-yellow or yellow-to-red).  The SNMP targets have been configured properly and when I generate an alarm (such as lost network redundancy) it usually sends a trap as expected, but every now and then it doesn't send anything.

The hardware status tab reports the failure correctly and the alarm always appears in vCenter within a few seconds, but the alarm doesn't always trigger the SNMP trap like it's supposed to.

Connectivity to the target is fine, it's pingable, and Wireshark confirms the SNMP packets were never sent, so it looks like vCenter has simply decided not to send them for some reason.

I've raised the alarms under various conditions to try identify the scenarios it does/doesn't send the trap but so far it appears to be completely random.  At first I thought it might have something to do with the frequency setting, even though it's configured to send an alarm once per state change (rather than repeat) I thought it might suppress them during what it may consider as a flapping scenario, but even seen it miss traps after leaving it idle for 10+ mins.

I'm doing most of my testing with 'lost network redundancy' and 'lost network connectivity' because they're the easiest to generate, however I've seen the problem with fan failures and storage failures too.

Any ideas?

Thanks,

Travis

[reish]

I'm seeing the same issue, but only from a yellow to a green clear. vCenter creates the event but never triggers an action which then tells SNMP to send it.

Most of our events are vm high cpu and so the Triggers were modified to

Trigger if any of the conditions are satisfied button set

VM CPU Usage (%)
Is above
90
for 5 min
100
for 60 min

Reporting Frequency 5 minutes

Actions set to

Send a notification trap
Once
Once
Once
Once

vCenter events

info           11/14/2011 1:49:33 PM                                                  Alarm 'Virtual machine cpu usage' on vm1 changed from Green to Yellow
info           11/14/2011 1:49:33 PM                                                  Alarm 'Virtual machine cpu usage' on vm1 triggered an action
info           11/14/2011 1:49:33 PM                                                  Alarm 'Virtual machine cpu usage': an SNMP trap for entity vm1 was sent
info           11/14/2011 1:49:53 PM                                                  Alarm 'Virtual machine cpu usage' on vm1 changed from Yellow to Green
info           11/14/2011 2:03:01 PM                                                  Task: Download patch definitions

[reish]

This was fixed by upgrading all the Hosts and vCenter to 4.1 U2 now U3. We haven't seen this issue since.