Help Center> Virtual Private Cloud> FAQ> Connectivity> How Do I Handle the IB Network Failure?
View PDF

How Do I Handle the IB Network Failure?

RDMA Communication Failure Between Two IB ECSs

  1. Check whether the Pkeys on the two ECSs are consistent.

    Run the following command to check for the Pkeys allocated to the ECSs:

    cat /sys/class/infiniband/mlx5_0/ports/1/pkeys/* | grep -v "0x0000"

    Figure 1 Checking Pkey consistency
    • If only one Pkey is obtained, contact technical support.
    • If two Pkeys are obtained, ensure that the two Pkeys on the two ECSs are the same.
  2. Run the following command to check whether the firewall is disabled:

    service firewalld status

    Figure 2 Checking the firewall

    If the firewall is not disabled, run the following command to disable it:

    service firewalld stop

  3. Check whether the RDMA communication command is correct.

    Run the following command on ECS 1 (client):

    ib_write_lat -x 0 --pkey_index 0 192.168.0.218

    Run the following command on ECS 2 (server):

    ib_write_lat -x 0 --pkey_index 0

No IP Address for the ECS IB Port

After you run the ifconfig command, it is found that no IP address has been assigned to the ECS InfiniBand (IB) port.

  1. Run the following command to check for the Pkey:

    cat /sys/class/infiniband/mlx5_0/ports/1/pkeys/* | grep -v "0x0000"

    Figure 3 Checking Pkey

    If only one Pkey is obtained, contact technical support.

  2. Run the following command to assign an IP address to the ECS IB port:

    dhclient ib0

    If no command output is displayed, the IP address cannot be obtained using DHCP.

  3. Contact technical support.
    After you have performed the preceding steps, if the IB network still cannot be used for communication or the IB port still cannot obtain an IP address, contact technical support for assistance and provide the technical support engineer with the following information.

    Item

    Description

    Example

    Your Value

    VPC1 ID

    VPC 1 ID

    Example: fef65559-c154-4229-afc4-9ad0314437ea

    N/A

    VM1 ID

    ID of ECS 1 in VPC 1

    Example: f7619b12-3683-4203-9271-f34f283cd740

    N/A

    VM2 ID

    ID of ECS 2 in VPC 1

    Example: f75df766-68aa-4ef3-a493-06cdc26ac37a

    N/A
Parent topic: Connectivity