Hi,
I had the same and other problems with SunCluster and solved them in the
following way:
Questions:
1) How to give different config files to the logical nodes ?
2) How to tell physical and logical nodes which file systems belong to
them ?
Response to 1)
I keep the config files for each tsm client like this:
In node1:
/opt/tivoli/tsm/client/ba/bin/node1/dsm.sys
/opt/tivoli/tsm/client/ba/bin/node1/dsm.opt
/opt/tivoli/tsm/client/ba/bin/node1/en_US ->
/opt/tivoli/tsm/client/ba/bin/en_US
/opt/tivoli/tsm/client/ba/bin/cluster1/dsm.sys
/opt/tivoli/tsm/client/ba/bin/cluster1/dsm.opt
/opt/tivoli/tsm/client/ba/bin/cluster1/en_US ->
/opt/tivoli/tsm/client/ba/bin/en_US
In node2:
/opt/tivoli/tsm/client/ba/bin/node2/dsm.sys
/opt/tivoli/tsm/client/ba/bin/node2/dsm.opt
/opt/tivoli/tsm/client/ba/bin/node2/en_US ->
/opt/tivoli/tsm/client/ba/bin/en_US
/opt/tivoli/tsm/client/ba/bin/cluster1/dsm.sys
/opt/tivoli/tsm/client/ba/bin/cluster1/dsm.opt
/opt/tivoli/tsm/client/ba/bin/cluster1/en_US ->
/opt/tivoli/tsm/client/ba/bin/en_US
NOTE: You can choose different names for dsm.opt and use the command line
option "-optionfile=FILENAME" with dsmc. However, since there is no command
line option to indicate which "dsm.sys" file to use, you may have a problem
if you want "dsm.sys" of node1 to be different from dsm.sys of cluster1.
That's why I prefer to keep the files in separet dirs.
Now,
To backup Node1, I do:
[root@node1] # . tsm_env node1 <<<---- see the dot at the begining
;-)
[root@node1] # dsmc inc
To backup cluster1 that is now running on *Node1*, I do:
[root@node1] # . tsm_env cluster1
[root@node1] # dsmc inc
or to backup cluster1 that is now running on *Node2*, I do:
[root@node2] # . tsm_env cluster1
[root@node2] # dsmc inc
The catch is that the "tsm_env" script changes the "DSM_CONFIG" and
"DSM_DIR" environment variables to point to the correct place:
____________________________________________________________________
#!/usr/bin/ksh
if [ '/usr/bin/tsm_env' = "$0" ] || [ 'tsm_env' = "$0" ] ; then
unset CLIENT
else
CLIENT=$1
fi ;
if [ "help" = ${CLIENT} ] || [ ! "${CLIENT}" ] ; then
echo "\nSYNOPSIS:\n\t . tsm_env { CLIENTE_NAME | help | clear }
\n\n"
elif [ "clear" = ${CLIENT} ] ; then
echo "Clearing TSM environment variables... \n"
unset DSM_DIR DSM_CONFIG
else
echo "\nSetting TSM environment variabables for client=\"${CLIENT}
\" \n";
DIR=/opt/tivoli/tsm/client/ba/bin/${CLIENT}/;
export DSM_DIR=${DIR};
export DSM_CONFIG=${DIR}dsm.opt;
fi
shift $#
____________________________________________________________________
Response to 2)
Suppose,
Node1 has 3 file systems: / , /opt and /var <<-- on local disks
Node2 has 4 file systems: /, /opt, /var and /export <<-- on local disks
Cluster1 has 2 file systems: /app/appl1, /app/data1 <<-- on shared
disks
In the Include/exclude files of *both* node1 and node 2, you add the
following tsm options:
exclude.fs /app/appl1
exclude.fs /app/data1
In the Cluster1 "dsm.sys" file you add:
domain /app/appl1
domain /app/data1
Now, what happens is this when you run a backup:
Node1: Will *ignore* the cluster's file systems ---> (/app/appl1
and /app/data1)
Node1: Will backup it's local file systems ---> (/, /opt and /var)
Node2: Will *ignore* the cluster's file systems ---> (/app/appl1
and /app/data1)
Node2: Will backup it's local file systems ---> (/, /opt/, /var and
/export)
Cluster1: Will *only* backup it's file systems ---> (/app/appl1
and /app/data1)
Hope this helps,
Justino Costa
justino.costa AT edinfor.edp DOT pt
Kurt Beyers <kurt.beyers AT DOLMEN DOT BE>@VM.MARIST.EDU> on 10-05-2001 13:47:09
Please respond to kurt.beyers AT dolmen DOT be
Sent by: "ADSM: Dist Stor Manager" <ADSM-L AT VM.MARIST DOT EDU>
To: ADSM-L AT VM.MARIST DOT EDU
cc:
Subject: TSM and MC/Serviceguard Cluster on HP-UX
Hi,
I need to backup MC/SG packages (the cluster packet of HP-UX) using TSM
(TSM
3.7 running on a NT server). As far as I understand, TSM can't make a
cluster aware backup (define the package as a client using it's IP
address).
This causes a few problems. For example, if the package switches from node,
the next incremental backup will be a full one because TSM sees the
filesystems of the package as new filesystems of the backup node. The same
for a restore, you should check first on which node the package is running
and I guess there will be a few other problems I haven't thought about yet.
How do you use TSM to backup a HP-UX cluster, are you using
workarounds/scripts for the problems? Are there still other problems with
the combination TSM/Serviceguard?
Any input would be greatly appreciated. Thanx,
Kurt
|