Serviceguard Extended Distance Cluster (EDC) with VxVM/CVM Mirroring on HP-UX, May 2008

Administrative VxVM/CVM tasks in an EDC
Taking a site offline
The main purpose of an EDC is to provide a highly available environment which can continue to
provide computing resources to applications even if an entire data center becomes unavailable, either
due to a disaster or scheduled maintenance.
The following procedure should be used to take a complete data center offline for planned
maintenance:
• All application packages that run on nodes in the data center that you want to take offline must be
failed over to an adoptive node in the data center that will stay online.
• Halt all of the cluster nodes in the data center that you want to take offline, after all application
packages have stopped running on them.
• Break off the plex associated with each volume that resides on the storage system in the data center
that you want to take offline. Using the example shown in the vxprint output from figure 7, the
following command will prepare the volume vol1 for a shutdown of data center DC2:
vxsnap -g dgEDC make source=vol1/newvol=SNAP-vol1/plex=vol1-02
It is important, to identify the plexes associated with each volume that resides in DC2. In this
example, the plex vol1-02 is located on subdisk DC2-EVA8k1_6-01. All plexes located in DC2
need to be broken off from their volumes in this way.
VxVM/CVM 5.0 Tip:
With the new “site-awareness” feature all plexes of all volumes that reside
in a specific site (data center) can be detached in a single operation per
disk group.
vxdg -g dgEDC [-f] detachsite DC2
From this point forward, the data in site DC2 will not be updated. Any data update is done only to
plexes in DC1 and is recorded in the FR maps of the associated DCL volumes. Recording the changes
in the FR maps provides for delta synchronization at the time the secondary site is brought up. If DC1
fails before DC2 is brought up, only old data will be available at DC2.
To bring a site back online, corresponding steps to those mentioned above should be performed in
the reverse order. The following command re-attaches the snapped volume SNAP-vol1 back to
vol1: “vxsnap -g dgEDC reattach SNAP-vol1 source=vol1“
Since the changes to volume vol1 were tracked in the FastResync map while one plex was broken off,
only the changed regions need to be synchronized.
VxVM/CVM 5.0 Tip:
With the new “site-awareness” feature a single operation re-attaches all
detached plexes to their volumes within a disk group.
vxdg -g dgEDC [-o overridessb] reattachsite DC2
16