I posted a query to the list a few weeks ago asking if anyone else had
seen problems with offline catalog backups taking excessive amounts of
time a Sun T2000 running Solaris 10 and with NetBackup 6.0MP4. Several
people e-mailed me informing me that Veritas recommends using the hot
catalog backup functionality instead of offline catalog backups, which I
already knew and which was completely unhelpful as I was concerned about
why an offline catalog backup that used to take about an hour and 15
minutes was now taking over five hours without any increase in the size
of the catalog. Ignoring this sort of drastic change in a system's
behavior is incredibly bad practice and I wanted to get to the bottom of
this and not just assume that because I could run hot catalog backups
that everything was going to be OK.
 
I did get one very helpful e-mail from Don Klebba who had had a similar
problem. Don found that disabling TCP fusion and changing some of his
TCP parameters with ndd fixed the problem. Armed with the information
that Don gave me (Thanks Don!) I found the following Symantec tech note.
 
http://seer.entsupport.symantec.com/docs/284421.htm
 
which explains the TCP fusion bug and how to fix it; you can either
install the latest Sun patch bundle, which also fixes the T2000's
problem with crashing when jumbo frames are enabled on the network
interfaces, or disable TCP fusion with mdb and then edit your
/etc/system file so that on the next reboot your system will no longer
be using TCP fusion. 
 
If it hadn't been for Don's e-mail I wouldn't have found this because
the tech note is about catalog backups failing with type 11 (a system
call failed) errors. My catalog backups weren't failing, they had just
become incredibly slow. I was able to take the downtime to install the
patch bundle and my offline catalog backups went back to taking an hour
and 15 minutes. Again I'd like to say thanks to Don Klebba and point out
that when this sort of thing happens it's a good idea to figure out why
it's happening and not just assume that because you have an alternative
solution (in this case hot catalog backups) that things are going to be
OK. I've seen some system administrators get bitten because they made
that sort of assumption and end up having their careers suddenly
interrupted with an RGE (resume generating event).


_______________________________________________
Veritas-bu maillist  -  Veritas-bu AT mailman.eng.auburn DOT edu
http://mailman.eng.auburn.edu/mailman/listinfo/veritas-bu