HPE HP 5900AF-48XG-4QSFP+ behaving weirdly with multiple SNMP v3 pollers

[jays7110]

Hi,

I have an HP comware switch HP 5900AF-48XG-4QSFP+ in my network on Version 7.1.045, Release 2307 which is being poller by SNMP servers using SNMPv3. Switch works fine with 2x SNMP servers but as soon as a 3rd one starts polling, the switch starts behaving temperamentally and stops responding to servers intermittently. i.e. it will stop respinding to all servers intermittently as if its got a bottleneck somewhere. Once the 3rd SNMP server one is stopped from polling, it goes back to performing normally. 
CPU and memory stats are nominal, no massive or sudden increases in CPU or memory utilization before during or after the 3rd SNMP server polling. The interfaces are also at high utilization.

CPU around 10% throughout, memory around 72% free throughout.

Any idea what this could be?

[Ivan_B]

Hi @jays7110 !

All Comware-based switches have internal rate-limiters protecting CPU from overwhelming traffic. As far as I know 5900 have limit of 500 packets per second. You can check if your 3 servers do not pass that limit by executing following commands:

system-view
probe
debug rxtx softcar show slot <slot>

Instead of <slot> insert your IRF slot number, if it is '1', the command will be 'debug rxtx softcar show slot 1'

SNMP traffic has category number 63. Check the column named 'DisPkt_All' it represents dropped excessive traffic.

If the counter is '0' or doesn't increase when SNMP is stalled, then it may be a bug and I really suggest you to upgrade your switch, as the version it is running on is really old.

Hope this helps!

[jays7110]

Thanks Ivan. I will look into it and update this post with the results.

kind regards