To enable concurrent maintenance, configure IBM Spectrum Virtualize™ nodes in pairs. If one IBM Spectrum Virtualize node is being serviced, the other node
can keep the SAN operational.
With concurrent maintenance, the hardware on one
IBM Spectrum Virtualize node can be serviced or replaced
while the SAN and host systems are doing productive work.
Attention: Do not remove the power from both IBM Spectrum Virtualize nodes unless the procedures instruct
you to do so.
Verify that concurrent maintenance is enabled before you shut down a node that is part
of a system or when you delete the node from a system. To do so, complete the following checks.
- Confirm that no volumes have dependencies on the node.
In the management GUI, select . Right-click the node that you might need to shut down or delete from the system to
show a list of actions for that node. Click Show Dependent Volumes to display
all the volumes that depend on a node. You can also use the node parameter with the
lsdependentvdisks CLI command to view dependent volumes.
If dependent volumes exist, determine whether the volumes are being
used. If the volumes are being used, either restore the redundant configuration or suspend the host
application. If a dependent quorum disk is reported, repair the access to the quorum disk or modify
the quorum disk configuration.
- Ensure that the host multipathing device drivers can fail over to the partner node.
Some host
multipathing device drivers take a while to update after changes are made on the fabric. Do not shut
down a node or delete the node from a cluster if the partner node in the I/O group to which the node
belongs has not been online for more than 30 minutes.
If possible, check the status of the
host multipathing device drivers before shutting down a node to ensure that the device drivers can
fail over to the partner node.
When you shut down the node, follow the procedure that is described in MAP 5350: Powering off a node.
When you delete a node from the clustered system, retain the node information that is described in
Deleting a node from a clustered system by using the management GUI. This information will help you
avoid data corruption when you add the node back to the system. The topic describes how to ensure
that the multipathing device driver does not rediscover any paths that are manually removed. Other
considerations about dependent volumes are also provided.
For more information about working with dependent volumes, see the following topics: