[S5731] SNMP polling problem

I also tried:

set snmp dynamic-control threshold 90

snmp-agent packet max-size 17940
snmp-agent packet-priority snmp 7
snmp-agent protocol server message queue 100

Didn't help either.

Hi eks41,
If the switch responds slowly or times out to the SNMP NMS, perform the following measures:

Solution 1: Multiple terminals may access the S series switch at the same time. The SNMP module of the switch needs to respond to multiple terminal users concurrently, resulting in a delay. In this case, you can run the snmp-agent acl command to configure an SNMP ACL to restrict the access of other users.

For example, perform the following configuration:

# Allow only the NMS that matches ACL 2000 to access the switch through SNMP.

<HUAWEI> system-view
[HUAWEI] acl 2000
[HUAWEI-basic-2000] rule permit source 192.168.10.10 0
[HUAWEI-basic-2000] quit
[HUAWEI] snmp-agent acl 2000

Solution 2: On the SNMP NMS, change the SNMP request sending interval to 1 minute or change the resending timeout interval (currently 0.5s).

Solution 3: Try to install patch V200R019SPH012.

I hope it helps!

Hi DDSN,

3. Huawei Support portal does not have V200R019SPH012 in the list of available patches for my S5731, just V200R019SPH029 (the latest, which I have installed), V200R019SPH020 and V200R019SPH007.

1 & 2. We use NMS for SNMP-polling all network devices every 10 seconds, no other devices send SNMP requests.

It is an industrial environment with strict rules, so we can't increase polling interval.

But for testing I tried different NMS polling params - it didn't help.

The problem is not delaying, 1 or 2 seconds pause for SNMP answer is acceptable. The problem is that S5731 very often stops in the middle of its answer for SNMP request until the next request no matter of NMS polling or timeout interval.

All other our network devices (Huawei AR6140 routers, Cisco IE-2000 switches, PLCs, workstations) polled fine with the same params.

LaiLai18,

I'd be happy if solution were that simple :)

In fact, link is perfect. NMS server connected directly to the S5731-S.

And the switch responds but very often stops in the middle of responding.

I continue my research on the problem.

Looks like S5731's SNMP engine has a serious defect. With NMS off I even get timeout errors sometimes while doing SNMP-Walk requests with a MIB browser. And with NMS on SNMP-Walk requests by MIB browser mostly finish with timeout errors after receiving part of some table.

Screenshot attached.

DDSN, thank you, CPCAR was a key!

Solution:

cpu-defend policy allow-snmp

 car packet-type snmp cir 768

cpu-defend-policy allow-snmp global

cpu-defend-policy allow-snmp

Default snmp cir value was 128.

CPU load raised from about 16% to about 21%, not critical. NMS now works fine.

Thanks again!