Help Center/ Cloud Container Engine/ User Guide/ Clusters/ Upgrading a Cluster/ Troubleshooting for Pre-upgrade Check Exceptions/ Nodes' System Parameters
Updated on 2026-01-04 GMT+08:00
View PDF
Share

Nodes' System Parameters

Check Items

Checks whether the system parameters on your node meet the expectation.

Solutions

  • Scenario 1: If the MTU value of the bond0 network on your BMS node is not the default value 1500, this check item will fail.

    Non-default parameter settings may lead to service packet loss. Change them back to the default values. Take the following steps:

    1. Disable bond0. 
      sudo ip link set dev bond0 down

      Disabling bond0 will interrupt services. Exercise caution when performing this operation.

    2. Change the MTU value to 1500. 
      sudo ip link set dev bond0 mtu 1500
    3. Restart bond0. 
      sudo ip link set dev bond0 up
    4. Verify the modification result. (Ensure that the MTU has been updated.) 
      ip link show dev bond0 | grep mtu

      If the command output is mtu 1500, the setting has been modified.

  • Scenario 2: If you upgrade the cluster to v1.29 or later, the kernel parameter nf_contrack_max is set by kube-proxy. If the value of this parameter is greater than that of conntrack-min configured for kube-proxy, this check item will fail.
    You are advised to use the kube-proxy startup parameter conntrack-min to ensure that the kernel parameter value is not less than the value configured on the current node after the upgrade.
    1. View the nf_contrack_max kernel parameter configuration. 
      cat /proc/sys/net/netfilter/nf_conntrack_max
    2. Modify the conntrack-min of kube-proxy through node pool configuration management. Set it to the value of the nf_contrack_max kernel parameter on the current node (which is the value queried in the previous step).

Parent Topic: Troubleshooting for Pre-upgrade Check Exceptions