Backing up the cluster configuration
This topic describes the overall
procedure for backing up a cluster configuration and the conditions that must
be satisfied in order to perform a successful backup.
Plan to regularly back up the business data that is stored on
all VDisks using your preferred backup method. It is important that the data
on all VDisks is backed up, because it will be lost when the configuration
is restored.
Prerequisites
- All nodes must be on-line.
- No object name may begin with an underscore.
- All objects should have non-default names, that is, names that
are not assigned by the SAN Volume Controller.
 | Note: While it is recommended that
objects have non-default names at the time the backup is taken, this is not
mandatory. Objects with default names will be renamed when they are restored.
The names will appear in the format name_r. |
Context
The following scenario illustrates the value of configuration
backup:
-
Use the Backing up a Cluster Configuration
panel in the master console or the svcconfig command to create
a backup file on the cluster that contains details of the current cluster
configuration.
- Store the backup configuration on some form of tertiary storage.
 | Note: You must copy the backup file off the cluster or it will be lost
if the cluster crashes. |
- If a severe failure occurs, it will cause the cluster to be lost. Both
configuration data (for example, the cluster definitions of hosts, I/O groups,
managed disk groups, MDisks) and the application data on the virtualized disks
is lost. In this scenario, it is assumed that the application data can be
restored from normal customer backup procedures. However, before this can
be carried out, it is necessary to reinstate the cluster, as configured at
the time of the failure. This means restoring the same managed disk groups,
I/O groups, host definitions, and finally the VDisks that existed prior to
the failure. The application data can then be copied back onto these VDisks
and operations resumed.
- Recover the hardware: hosts, SAN Volume Controllers, disk controller systems, disks, and
SAN fabric. The hardware and SAN fabric must physically be the same as that
used before the failure.
- Reinitialize the cluster.
- Restore your cluster configuration using the backup configuration file
generated in step 1
- Restore the data on your virtual disks (VDisks) using your preferred restore
solution, or with help from IBM Service.
- Resume normal operations.
Restrictions: The following restrictions must be
observed for the scenario described at *** or a
similar scenario to work:
- The installed hardware must be identical when the backup is
taken and when the restore is done. Otherwise the restore will fail.
- No changes should be made to the fabric or the cluster between backup
and restore. If changes are made, you should back up your cluster configuration
again.
- There are two phases in the restore process, prepare and execute. Do not
make any changes to the fabric or the cluster between the two phases.
- No independent operations that could change the cluster configuration
should be running while the backup command is running.
Steps
Perform the following steps to backup your cluster configuration:
-
Back up the data that your enterprise
uses to run its business using your preferred backup method. It is important
that the data on all VDisks is backed up, because it will be lost when the
configuration is restored.
- Back up the cluster configuration using the Backing Up a Cluster Configuration
panel or the svcconfig backup command.
 | Note: Back up the cluster configuration immediately after completing step 1. |
- Ensure that all your SSH keys are available. You need these keys when
you restore the cluster configuration.
Result
When the svcconfig backup command runs, it produces
a file called svc.config.backup.xml, which describes the cluster configuration.
This file is stored in /tmp on the configuration node within the cluster.
It is important that this file is copied from the cluster to some external
storage since, should the configuration node move to another node within the
cluster, then the /tmp directory on this node will be inaccessible. (The
configuration node might move in response to an error recovery action, or
due to some user maintenance activity.)
To copy the svc.config.backup.xml from the node to external storage, use
the secure copy command (pscp) on the master console or the secure copy command
as described in the example below if you accessing the cluster through your
own secure shell installation.
Example
The backup feature of the svcconfig command is designed
to back up the cluster information, such as VDisks, local Remote Copy information,
MDiskgrps, and nodes and not the information that you have written to the
VDisks within the cluster. It is important that any application using the
VDisks on the cluster as storage should back up its data as usual using their
appropriate backup routines.
All nodes must be online and no object in the cluster can begin with an
underscore "_".
To create a backup of your cluster configuration, (your_cluster_name),
perform the following steps:
- Log onto the cluster by issuing the following:
ssh -l admin your_cluster_name -p 22
This will bring up a session on the cluster where you can issue
the svcconfig command.
- Issue the following:
svcconfig clear -all
This
will remove any existing backup files that are on your cluster and ensure
a clean directory into which the backup files can be placed.
- Issue the following:
svcconfig backup
The cluster
will return output similar to the following as the backup runs:
CMMVC6112W mdisk mdisk14 ...
CMMVC6112W node node1 ...
CMMVC6112W node node2 ...
Once the backup
has completed and you have returned to the prompt, you will need to exit the
cluster and copy the backup files to somewhere off the cluster.
- Issue the following:
exit
- To copy the backup files off the cluster, issue the following:
scp -P 22 admin@your_cluster:/tmp/svc.config.backup.*
/offclusterstorage/
The following three files will be retrieved
from the cluster:
- svc.config.backup.xml This contains the information
about the objects on your cluster.
- svc.config.backup.sh This contains the svcinfo commands that were used to create the backup of the cluster.
- svc.config.backup.log This contains the feed back
from the backup routine and will contain any error information that may have
been reported.
Because the svcconfig command requires that the
same cluster configuration be in place before you can restore, it is wise
to rename these files with the configuration node name either at the start
or end of the file names to make identification easier at restore time. To
rename the files with the configuration node name, issue the following:
mv /offclusterstorage/svc.config.backup.xml
/offclusterstorage/svc.config.backup.xml_myconfignode
Because these files contain details about your cluster, it is advisable
to copy them to a location that is under password control to avoid unauthorized
access to this configuration information.
Related topics