i have some 3750E switchstacks and use snmp to monitor the bandwidthusage. Sending the snmp-requests - in fact I use the oid's 188.8.131.52.184.108.40.206.101XX and 220.127.116.11.18.104.22.168.101XX (in/out octets) - to the stackmaster works fine and a clear graph is displayed. Monitoring a port on a stackmember (e.g. oid 22.214.171.124.126.96.36.199.10615 for port 15 on switch 2) results in peaks like this:
First I assumed that the switch needs to much time to answer the request, but the cpu usage is normal (~5-10%) and a snmp-manual request is answered instantly. Some more manual requests revealed that the octet-counter is not updatet for seconds!
Next thought: Maybe the switch is faulty. So i replaced it with a reserve switch. Guess what - the same problem...
Then i tried to swap the stackroles. Now switch 1 is the stackmember and switch 2 is the stackmaster. The counters on switch 2 are updated in realtime, the counters on switch 1 are updated in seconds! What's that????
Has anybody an idea to update the switchmember-counters like them of switchmaster?