> I had to change the IP address of our master server/ACSLS/SN6000 to
> move them to another subnet last week.  The host names did not
> change.  Before the change, we were having no problems, but now, on
> 3 of my media servers, I have backups failing with error 23, socket
> read error.  In the bpbrm log on the media server, I'm seeing a lot
> of "broken pipe" error messages, and messages "cannot write FILE
> ADDED message to stderr".  There is no consistency in when the error
> happens.  Sometimes the backup starts and runs for a while before it
> fails, other times it never starts at all.  Some backups on the
> media server complete with no error.

Inconsistent network errors are almost always due to actual packet
errors in my experience.  Try netstat -Ian hme0 (on Sun, been too long
since I've been around SGI).

I'd guess that when the net connections were moved to switch ports
which are set either wrong or to auto.

> 
> One of the media servers is running IRIX, and the other two having
> the problem are Solaris.
> 
> Another bit of strange behavior I'm seeing is that when running a
> backup manually, it takes a long time before it even shows up in the
> job monitor as "queued".  Sometimes as long as 20 minutes.

Use netstat -an | grep SYN to find out which client or media server
it's trying to connect to.  See also bpsched log.  bpsched connects to
clients to get filesystem lists (if using multiple data streams) and
to media servers to get available drives during this period.
-- 
Larry Kingery 
         "Cut down a tree with a herring? It can't be done!"