XGS4600-32F 4.50(ABBI.1) - High CPU usage when querying SNMP zynos_zyxelTransceiverDdmiTable

CppMonkey
CppMonkey Posts: 1
First Anniversary
edited August 2022 in Switch
Hi,

I've developed an OS module for LibreNMS. Currently it is able to poll :-
  • Stack CPU load
  • Stack memory utilisation
  • Unit temps
  • Unit fan speeds
  • Unit voltages
But as soon as we started polling the optic DDMI table, the CPU jumped to 100% every 5 minute (LibreNMS polls... every 5 minutes)

My LibreNMS repo branch "zynos-transcievers" in located on GitHub

The difference between the Zynos and Zynos-transcievers

Perhaps this is related to the DDMI issue we've experienced on 4.40(ABBI.2). I'll update a spare stack to 4.60(ABBI.0).

The units are run idle at 17-20% with 56 transceivers in 2 units. This is in sharp contrast to the XGS4600-52F running at 100% with 47 transceivers.

For the 32F units, the CPU utilisation doesn't change, even with 116 live transceivers.


The issue is when we try to query the DDMI tables via SNMP to monitor optics for potential issues.

Regards,
Paul 

All Replies

  • Zyxel_Derrick
    Zyxel_Derrick Posts: 126  Zyxel Employee
    First Anniversary Friend Collector First Answer First Comment

    Based on your topic, we assume the problem is on XGS4600-32F
    However, after reading the description, we are a bit confused 
    Therefore, we would like to clarify the info below

    1. "The units are run idle at 17-20% with 56 tranceivers in two units"
       Do you mean when XGS4600-32F was stacking and upgraded to firmware 4.60, the cpu high did not happen when polling DDMI table, am I right?

    2. "This is in sharp contrast to the XGS4600-52F running at 100% with 47 tranceivers"
       Do you mean when XGS4600-52F was stacking and upgraded to firmware 4.60, the cpu high happened when polling DDMI table, am I right?

    3. "For the 32F units, the CPU utilisation doesn't change, even with 116 live transceivers"
       Based on this description, we assume the problem is on XGS4600-52F, am I right?

    4. May I know if the test scenario of both 32F and 52F are the same?
       Also, all transceivers are connected with fiber cable and they are link up?
       If no, the reason that causes CPU high is that our switch will scan transceivers to make sure its link speed.
       If transceivers do not connect with cable, switch will keep pollingand lead to CPU high.
       

    Thanks for your help in advance
    Zyxel_Derrick