Sun Microsystems
Products & Services
 
Support & Training
 
 

Previous Previous     Contents     Index     Next Next

A Master Node Is Not Elected at Startup

At startup, the first master-eligible node that you boot should become the master node. The second master-eligible node that you boot should become the vice-master node. If the first master-eligible node does not become the master node, perform the following procedure.

ProcedureTo Investigate Why a Master Node Is Not Elected at Startup

  1. Log in to the first master-eligible node as superuser.

  2. Confirm that the /etc/opt/SUNWcgha/not_configured file does not exist.

    • If the file does not exist, go to Step 3.

    • If the file exists, delete it and reboot the node:

      # init 6

  3. Confirm that the target.conf file has the attribute flag set to "-".

    For more information, see the target.conf(4) man page.

    This attribute flag indicates that a master-eligible node is qualified to become the master node. The target.conf file contains the node description saved by the nhcmmd daemon on the master node. When a master node exists, or when the cluster is running, do not edit the target.conf file.

    • If the attribute flag is set to "-", go to Step 4.

    • If the attribute flag is not set to "-", do the following:

    1. Go in to single user mode:

      # init s

    2. Edit the target.conf file to set the attribute flag to "-".

      The node can be set with more than one attribute flag. Make sure that the flag "-" is the only flag that is set.

    3. Reboot the node:

      # init 6

  4. Confirm that the node has write access to the cluster_nodes_table file.

    For more information, see the cluster_nodes_table(4) man page.

    • If the node has write access to the file, go to Step 5.

    • If the node does not have write access to the file, do the following:

    1. Change the access permissions as described in the chmod(1) man page.

    2. Reboot the node:

      # init 6

  5. In the cluster_nodes_table file, confirm that the node attribute flag is set to "-".

    The "-" attribute flag indicates that the node is qualified to become the master node.

    • If the attribute flag is set to "-", go to Step 6.

    • If the attribute flag is not set to "-", do the following:

    1. Go in to single user mode:

      # init s

    2. Edit the cluster_nodes_table file to set the flag to "-".

      The attribute can be set with more than one attribute flag. Confirm that the attribute flag "-" is the only flag that is set.

    3. Reboot the node:

      # init 6

  6. If you cannot resolve this problem, contact your customer support center.

Two Master Nodes Are Elected at Startup

At startup, when the first master-eligible node becomes the master node, the second master-eligible node should become the vice-master node. If the second master-eligible node cannot detect the master node, it will become the master node. The presence of two master nodes is an error scenario called split brain.

A direct link between the master-eligible nodes prevents the occurrence of split brain when the communication between the master node and vice-master node fails. For information about the direct link, see the Netra High Availability Suite Foundation Services 2.1 6/03 Overview.

If your cluster is configured to use a direct link, perform the procedure in To Investigate Split Brain on Clusters With a Direct Link. If your cluster is not configured to use a direct link, perform the procedure in To Investigate Split Brain on Clusters Without a Direct Link.

ProcedureTo Investigate Split Brain on Clusters With a Direct Link

  1. Confirm that the direct link is physically connected to the serial ports of both master-eligible nodes.

  2. Confirm that the nhfs.conf file contains the following parameters:

    Cluster.Direct-Link.Backend=serial
    Cluster.Direct-Link.Heartbeat=20
    Node.Direct-Link.serial.Device=/dev/term/b 
    Node.Direct-Link.serial.Speed=115200

    The Cluster.Direct-Link.Heartbeat can have values up to one second.

    The Node.Direct-Link.serial.Speed can have one of the following values: 38400, 57600, 76800, or 115200.

  3. If the direct link is connected and configured correctly, and you still have a split brain error, contact your customer support center.

ProcedureTo Investigate Split Brain on Clusters Without a Direct Link

  1. Access the consoles of the master nodes.

  2. Confirm that you have two master nodes.

    On the console of each master-eligible node, run:

    # nhcmmstat -c all

    Each master node should see itself as master, and see the other master as being out of the cluster.

  3. Test the communication between the master nodes.

    On the console of each master-eligible node, run:

    # nhadm check starting

    When this command is run on a node, the command pings all of the other nodes in the cluster. If one eligible node cannot ping the other eligible node, the nodes are not communicating.

    If the Carrier Grade Transport Protocol (CGTP) is installed, the nhadm check command pings both of the network interfaces and the CGTP interface. If CGTP is not installed, the nhadm check command pings one network interface only.

  4. Evaluate the result obtained in Step 3 by using the following table.

    Table 4-1 Results of nhadm check starting Run When Two Master Nodes Are Elected at Startup

    Results of nhadm check

    Possible Cause

    Action

    Two network interface cards (NICs) fail, or one NIC fails and one NIC passes

    Incorrect switch configuration or incorrect cabling

    Reconfigure the hardware as described in the Netra High Availability Suite Foundation Services 2.1 6/03 Hardware Guide.

    Two NICs pass but the CGTP interface fails

    Incorrect Foundation Services configuration

    Examine the nhfs.conf and cluster_nodes_table files.

    Two NICs and the CGTP interface pass

    The master-eligible nodes exist in different domains.

    Confirm that the nodes have the same values for the domainid parameter in the nhfs.conf file.

  5. Confirm that all of the packages and patches are installed.

    1. Access the consoles of the master-eligible nodes.

    2. Display the installed packages and patches:

      # patchadd -p
      # pkginfo

    3. Compare the list of installed packages and patches with the lists defined in the Netra High Availability Suite Foundation Services 2.1 6/03 README and patch READMEs.

      • If a required package or patch is not installed on the master-eligible node, install it and reboot both master-eligible nodes.

      • If all of the required packages and patches are installed, go to Step 6.

  6. If you cannot resolve this problem, contact your customer support center.

Previous Previous     Contents     Index     Next Next