Shutting down a cluster or single node using the CLI Advanced function cluster overview using the CLI

This task provides step-by-step instructions for shutting down a cluster using the command-line interface (CLI).

Prerequisites

If all input power to a SAN Volume Controller cluster is to be removed for more than a few minutes, (for example, if the machine room power is to be shutdown for maintenance), it is important that the cluster is shutdown before the power is removed. The reason for this is that if the input power is removed from the uninterruptible power supply units without first shutting down the cluster and the uninterruptible power supplies, the uninterruptible power supply units will remain operational and eventually become drained of power.

When input power is restored to the uninterruptible power supplies they will start to recharge but the SAN Volume Controllers will not permit any I/O activity to be performed to the virtual disks until the uninterruptible power supply is charged enough to enable all the data on the SAN Volume Controller nodes to be saved in the event of an unexpected power loss. This might take as long as three hours. Shutting down the cluster prior to removing input power to the uninterruptible power supply units will prevent the battery power being drained and will make it possible for I/O activity to be resumed as soon as input power is restored.

Attention: Attention: Before shutting down a node or the cluster you should quiesce all I/O operations that are destined for this node or cluster. Failure to do so may result in failed I/O operations being reported to your host operating systems.
Attention: Attention: If you are shutting down the entire cluster, you will lose access to all VDisks being provided by this cluster.
Shutting down the cluster
Steps

Perform the following steps to shut down a cluster:

  1. Begin the process of quiescing all I/O to the cluster by stopping the applications on your hosts that are using the VDisks provided by the cluster.
    1. If you are unsure which hosts are using the VDisks provided by the cluster, follow the procedure called, Determining the hosts that a VDisk is mapped to.
    2. Repeat the previous step for all VDisks.
  2. Open a command prompt window.
  3. When all I/O has been stopped, issue the svctask stopcluster to shut down a single node or the entire cluster in a controller manner. If you specify the node ID or node name, you can shut down a single node.

    When you enter this command either a node ID or node name argument, the node in question is shut down. After the command completes, the other node in the I/O group destages the contents of its cache and goes into write-through mode until the power to the node is returned and the node rejoins the cluster.

    Attention: Attention: If this is the last node in an I/O group, you will lose all access to the virtual disks in the I/O group. Before you enter this command, ensure that this is what you want to do. You must specify the force flag.

    If a shutdown command has been sent to the cluster and both cluster and uninterruptible power supply units have powered off, when input power is restored it will be necessary to restart the uninterruptible power supply units by pressing the power button on the uninterruptible power supply front panel.

  4. Close the ssh session if you are using ssh in interactive mode.
Shutting down a single node
Attention: Attention: If you are shutting down a single node, and the other node in the I/O group is online, be aware that the cache on the partner node will go into write-through mode and that you are exposed to a single point of failure should the partner node fail while this node is shut down. Proceed to .
Attention: Attention: If you are shutting down a single node, and this is the last node in the I/O group, you will lose access to all VDisks being served by this I/O group.
Steps

Perform the following steps to shut down a single node:

  1. Begin the process of quiescing all I/O to the VDisks being served by this nodes I/O group.
    1. Determine the VDisks in question by requesting a filtered view of VDisks where the filter attribute is the I/O group in question. This can be done using the following command:
      svcinfo lsvdisk -filtervalue IO_group_name=<name>
      where <name> is the name of the I/O group in question.
    2. Once you have a list of VDisks, determine the hosts that these are mapped to by following the procedure called, Determining the hosts that a VDisk is mapped to.
  2. When all I/O has been stopped issue the following command to shut down the node:
    svctask stopcluster <nodename/ID>
    where <nodename/ID> is the name or ID of the node that you want to shut down.
    Note:Note: If this is the last node in the I/O group you also need to specify the -force parameter. For example to force the shutdown of node1:
    svctask stopcluster -force node1
Related topics Advanced function cluster overview using the CLI

(C) Copyright International Business Machines Corporation 2003, 2004