A question on bug CSCve35785 "Service "snmpd" hasn't caught signal 6 (core will be saved)"

[ryan.strayhan]

I read the symptoms and conditions of this bug, so I'm clear on that. High memory usage cause SNMP not to work. OK, then.

My question is why would diagnostics, such as SNMPWalk or 'Test Credentials' in my NMS system, work when actual polling does not? 

Funny thing is that with 3.1(3a), SNMPv1 worked, but our IT policy requires the Authentication of v3. Both v2 and v3 would pass diags and tests, but actual implementation failed. Working now, though, with 3.1(3b).

Thanks in advance for any insight.

[Kirk J]

Greetings.

Many of the SNMP issues that crop up tend to impact after many repeated polls at frequent intervals.

Doing a couple of walks and sample OID polls may not trigger the high memory usage that many polls over time would.  Is that something that you are observing when you say 'implentation failed', in that once the polling is setup it only runs successfully for a short time, before not getting responses anymore?

V1 only tracks 32bit counters, so maybe the V2 and higher natively are going to require more overhead anyway for the 64bit counters and hits the mem leak type condition quicker.

As you have noted, this bug is listed as a fixed caveat in 3.13b release notes: http://www.cisco.com/c/en/us/td/docs/unified_computing/ucs/release/notes/CiscoUCSManager-RN-3-1.html#reference_kff_ycf_31b

Thanks,

Kirk...