I have successfully configured media server to be cluster service on
Compaq Tru64 unix with using an IP alias as media server name and master
server installed on WinNT
I was unable to "FORCE" a scheduled onbar backup to use the cluster
alias name, it's because the spawn user-directed job "picked-up" the
"hostname" the backup class called up this job is used as the cluseter
Therefore, I needed to add the "real" hostname for both cluster nodes on
the client list of Informix backup class to make
the spawned informix user-directed backup job working.m But I've found
lots of netbackup traffic and the netbackup performance is bad, it took
more than 4 hours to fill up a DLT7000 tape and the backup capcity shown
over 140G for one tape at activity monitor, it's rare !!!
I would like to emphasis that for OS backup classes, they are all
I could configure all the daemons needed as media server and client as
cluster alias name, I have defined the backup class on using the cluster
alias nameas media server and client.
For the "netbackup" service whichever side it failover to, the backup
still working fine and also using the same tape. I have only problem on
informix agent at "onbar" user-directed job, it using the "hostname" of
the "netbackup" running on, not the cluster alias name.
Is that any parameter I can use to force that "onbar" user-directed job
to use the name that I specified, Thank you so much.
(I tried INFXBSA_SERVER, it did not help, how about CLIENT_NAME)
Customer compromished to use work around, that is:
Install a fiber HBA on NT master server, use that as "robot control
host" and then configure the cluster nodes as normal media server, but
another problem that we came across for this work around solution.
I set up equipment for configurng NT as a NBU "robot control host" and a
Tru64 cluster nodes to be media server to share the Compaq tape library
Unfortunately, I was only able to configure one cluster node to share
the tape library with NT, the "ltid" device management process would not
start after I configured another cluster member host. For backup
operation with NT and one unix host, it working fine. The "tape
library" and "tape devices" files are common shared by cluster member
nodes, to avoid the problem device file contention, I have created
individual device files for them but the problem still existed.