Veritas-bu

[Veritas-bu] Error 54 Solaris client

2002-03-28 16:39:53
Subject: [Veritas-bu] Error 54 Solaris client
From: larry.kingery AT veritas DOT com (Larry Kingery)
Date: Thu, 28 Mar 2002 16:39:53 -0500 (EST)
Denis Petrov writes:
> Unfortunately Detailed Logging did not show "gethostbyaddr failed"  on the
> master server
> I will ask the client to setup detailed logging on their system.

It wouldn't be on the master.  That would cause like a 41 or 48 error.

The error message ("timed out connecting to client") for 54 could be
better.  You see, it didn't time out connecting to the client at all.
It connected to the client just fine, then timed out waiting for the
client to connect back to it once it was done with it's preprocessing
and all that and ready to go.

If you keep in mind that there is a connection from the *media* server
to the client and then another connection back from the client to the
media server and that 54 means the second connection was never
received it's easier to troubleshoot.

> Now  I am thinking that client running NIS, NIS++ or DNS for their host name
> resolution and it may be causing some errors. I will have to investigate
> this.
> 
> >The problem is probably one of resolution - netbackup tends to use hostname
> rather than >interface name and this can cause conflicts - and if the
> clients use other DNS servers this >has caused problems in the past.

This is not correct.  NBU *almost never* uses the hostname at all.  It
uses the results of the gethostby* commands - if it looks like it's
using a hostname that's a coincidence.

> 
> Original setup was exactly like that... Now I am thinking that it is name
> resolution issue and nothing more. But what puzzles me, It is not consistent

Actually, since it's not consistent name resolution is one of the
first things I'd consider.  With round robin usage of multiple
mappings, the chance to use different servers, caching, etc...

Check your name servers and make sure they're all in agreement
forwards and backwards.  Maybe 20% of the time the primary is timing
out so a secondary is used?

Another thing that can cause 54 is if the client gets the IP back
okay, but the route back is not static.  If there's more than one hop
back to the media server, this is almost guaranteed (not that it will
change, but that it *could*).

Also, check netstat/ifconfig/lanadmin for packet error counts.  Also
keep in mind that other things might be occurring on the network
during the backup (e.g. with something set to half duplex and a heavy
load, it's *possible* that a response could timeout on resends)

> 
> >Oh and check both server and client can see the Master
>  Yes no problem there, BTW In my setup Master is NBU media server and NBU
> server
> 
> Anyway I have some things to check now.
> Thank you for your help
> 
> Denis
> 
> _______________________________________________
> Veritas-bu maillist  -  Veritas-bu AT mailman.eng.auburn DOT edu
> http://mailman.eng.auburn.edu/mailman/listinfo/veritas-bu

-- 
Larry Kingery 
         "Trust me, I know what I'm doing."  -- Sledge Hammer

<Prev in Thread] Current Thread [Next in Thread>