I have two HP-UX hosts as clustered NW servers. There's also a dedicate storage node. They're all connected to a tape library L700new. Is that possible to back up the bootstrap to tape instead of some file device on the servers?
What version is that? I do remember I had to use this on active node due to some documented issues which were supposed to be fixed in 7.3SP3 (meaning in that version -s should no longer be required). NW server will then see local devices as local to backup server.
Hmm... I had 7.2.2 for short time on HPUX, but I can't remember how it was there. But it should still work. Of course, your script should take care library in such manner that you do jbconfig upon failover on active node - that should address your issue no matter on which node your package is running.
Since I used 'jbconfig -s virtual_host' to configure that tape library, the jukebox is recognised as a remote device and no bootstrap is backed up... So I had to set up a file device to achieve it. So I have to connect each NW server with a standalone device to back up the bootstrap?
1. the virtual service networker is running on server1 2. on server2, run jbconfig -s networker 3. three paths for each drive - two servers, one dedicate storage node
Then in the resource file, I have: rd=server1:/dev/rmt/... rd=server2:/dev/rmt/... rd=dsn:/dev/rmt/...
When I set server1 active, the resource file becomes like: rd=server1:/dev/rmt/... rd=server1:/dev/rmt/... rd=dsn:/dev/rmt/...
When I set server2 active, it's like: rd=server2:/dev/rmt/... rd=server2:/dev/rmt/... rd=dsn:/dev/rmt/...
In my case 3 libraries on each side and active nodes always run script which: a) on local node calls jbconfig -s package_name b) via ssh executes jbconfig -s package_name on remote node
Those devices which are local to node where package is running are always local. -s for local node I had to implement because of documented issue in NW - I think with 7.2.2 everything is fine so try jbconfig without -s for local node - that should create devices local to active node.
Your server should not be shown as physical name, but rather as package name - that's one of the features your get with metro cluster. Perhaps something is wrong with your configuration files. Make sure you did follow instructions for cluster installation of NW step by step.
I didn't state correctly. When running nwadmin, the server is the package name, but the user shown on the top right conner is root@server1. And when using jbconfig -s package_name, it says '.... name must match'. I'm wondering if that's why.
ble1
6 Operator
•
14354 Posts
•
56186 Points
333
0
Posted September 7th, 2007 00:00