Prints Lost in Failover

[Malcolm McKenzie]

We have a two node Serviceguard configuration and have found that during a failover situation, prints are lost, what is the best way to avoid this?

[melvyn burnard]

You would need to use a Print Spoller that can be setup as a package within ServiceGuard is the best answer.
I have heard of people doing this with the standard lp spoller on HP-UX as well.

[Robin Wakefield]

Hi Malcolm,

Assuming your printer definitions are identical on both systems, I'd consider copying your /var/spool/lp directory to a directory on your shared vg (maybe created an lv just for this purpose), and linking /var/spool/lp to it.

For completeness you may want to create a service definition for the spooler, but I'm not sure if this would *definitely* be required.

Rgds, Robin.

[John Poff]

I've done a similar thing before (in HACMP on AIX). You just need to figure out which directories you need (/var/spool/lp, etc.), create a volume group with them as filesystems, and put them in a package. It means that your failover box will have to have print services stopped, or will have to be configured in the package to stop and restart print services when the printing package comes up on that node.

It's can be done, and if your printing is critical it is pretty cool.

JP

[A. Clay Stephenson · Accepted solution]

Hi Malcolm:

A little bit of clarification. Is this an Active-Active cluster so that both nodes have printing going on in normal circumstances? If so, this makes the task more difficult to use the standard lp spooler.

Clay

[Malcolm McKenzie]

Printing is only active on one server, i.e. the server where the application package is running.

[A. Clay Stephenson]

In that case, this is pretty simple. As has been mentioned, make /var/spool/lp a filesystem within your packages volume group.
Part of the package start/stop will be lpsched/lpshut. Note than /var/spool/lp/sinterface is symbolically linked to /etc/lp/interface. The same is true for /var/spool/lp/rsmodel. Make sure that those files are identical in /etc/ on both nodes and that should fix you. I assume that you have JetDirect installed, it will need to be installed in identical locations on both machines and any filters or programs which are called by your interface programs will also need to be installed in identical locations on both boxes. Don't forget to disable starting/stopping lp in the /sbin/init.d/scripts.

Regards, Clay