[Veritas-bu] Red Hat 9 Backups

Hi,
I run James' script on one of my Redhat 9 box, the system had no fix update,
jnbSA worked find but I still got 174 error when I backup it. I ran bpcd, I
got this error messages:

Incorrectly built binary which accesses errno, h_errno or _res directly.
Needs to be fixed.
setsockopt: Socket operation on non-socket

Then I upgrade the most current kernel and glibc.
kernel-2.4.20-24.9
glibc-2.3.2-27.9.7
The backup works find.

Thanks James, you rock.

Koping

Koping Wang
ESRI
Backup/Systems Administrator
380 New York St
Redlands CA, 92373



Message: 2
Date: Mon, 1 Dec 2003 17:42:07 -0600
From: James Chacon <jmc AT netbsd DOT org>
To: "Jon D. Benson" <jbenson AT neurome DOT com>
Cc: Scott Jacobson <SJACOBSO AT novell DOT com>,
   " (veritas-bu AT mailman.eng.auburn DOT edu)"
<veritas-bu AT mailman.eng.auburn DOT edu>
Subject: Re: [Veritas-bu] Red Hat 9 Backups

On Mon, Dec 01, 2003 at 03:28:10PM -0800, Jon D. Benson wrote:
> 
> 
> Scott Jacobson wrote:
> 
> >Red Hat Gurus,
> > 
> >I've seen many threads about the incompatibility of RH9 with V4.5 and 
> >the need to use the V3.4 client, however, can someone tell what are the 
> >actually characteristics of errors seen when a backup attempt is made?  
> >Error 174, 4x, 23....
> > 
> >Thanks in Advance,
> > 
> >Scott Jacobson
> >Novell, Inc.
> 
> The error type was 174. I opened a case with Veritas and they said this:
> "The reason you get the 174 errors is because NetBackup doesn't
> support RedHat 9 at this time.  I expect support to start as of Pack FP6
> which is due out at the end of October beginning of November timeframe.
> Since there is no workaround, I would like to close this case."
> 
> Any word on the release of this update? I checked last week but nothing.
> 

Ok, the reason one gets 174 errors is due to the data coming back to the
tape process isn't an even block size amount.

This is due to the following error that RH9 spits out on the veritas
binaries:

Incorrectly built binary which accesses errno or h_errno directly. Needs to
be fixed.

That gets sent across the pipe, which breaks the backups. The latest glibc
update for rh9 fixes that by putting the error on stderr but there are
other problems you'll run into there (such as bp not working correctly,
etc).

I used to have a wrapper which stripped this off the output but it was
clumsy and I found a better solution.

The solution:

Make sure all the binaries have LD_ASSUME_KERNEL=2.2.5 set in their
environment
before running them and everything works perfectly fine (due to the changes
in threading models between glibc releases).

So, after I install netbackup on a RH9 system (or actually RH AS/ES/WS 3.0
as it also suffers from this) I run the following: (yes there is more
efficient shell but this was quick)

# Fix for glibc-2.3 and threading
cd /usr/openv/netbackup/bin
rm bparchive
ln -s bpbackup+ bparchive+
for i in bp bpbackup bpbkar bpcd bpclimagelist bpclntcmd bpdynamicclient
bphdb bpjava-msvc bpjava-usvc bplist bpmount bprestore jbpSA jnbSA tar xbp
do 
mv $i $i+
echo '#!/bin/sh' > $i
echo 'LD_ASSUME_KERNEL=2.2.5' >> $i
echo 'export LD_ASSUME_KERNEL' >> $i
echo "exec /usr/openv/netbackup/bin/${i}+ \$*" >> $i
chmod +x $i
done
cat > bparchive <<"EOF"
#!/bin/sh
LD_ASSUME_KERNEL=2.2.5
export LD_ASSUME_KERNEL
exec /usr/openv/netbackup/bin/bparchive+ $*
EOF
chmod +x bparchive

James