I couple of weeks ago I was having trouble with amanda dumps for a
particular disk. The dump process would eat up lotsa CPU and could not
be killed (have to reboot). I was able to determine that the disk it
was backing up (using gnutar) was host:/home1 but in fact home1 was a
symlink home1-->home, so I changed home1 to home in the disklist and
everything ran fine.
Well, last night I decided to back up a Windows (2000) box for the first
time, using the amanda server as the samba client. I just tried a
//winhost/c$ share. Well, the same thing happened as with the home1
disk. The dump used all the CPU it could get and I had to reboot. I
looked at the /tmp/amanda files and nothing seemed unusual or out of
place. The amstatus command just says that it is dumping (0%). The
netstat command shows that the RevQ and SendQ of the amanda server
processes (where it is connected to itself) are rather large, and not
appearing to move. I dunno, it could be a kernel issue and not an
amanda issue. I'm just wondering if anyone else has experienced this.
my server:
Dell PowerEdge 4400 2GB RAM, 2 1GB Xeon CPUs, 410GB disk space
Red Hat Linux 7.2 w/ kernel-enterprise-2.4.9-13
Amanda: amanda-2.4.2p2-4
Holding disk is on a ReiserFS filesystem
DLT 4 35/70GB tapes/drive
All other disks (28 on 6 systems) back up fine. I do daily level 0's on
a 15-tape rotation.
--
--------------------------------------------------------------------------
Systems Engineering Better by far you should forget and
Dynacare Laboratories smile than that you should remember
www.dynacare.com and be sad.
-- Christina
Rossetti
