I am evaluating Veritas Netbackup on a Sun E220R attached to an HP Surestore
E440. The
library is scsi attached containing two Ultrium 230 drives. The scsi card
is a dual port pci
card with one port going to an Ultrium drive, then daisy chained to the
robot. The second
scsi port goes to the second Ultrium drive.
The Sun box is running Solaris 7 with the recommended patches. I have
applied the Veritas
jumbo patch J0850645.solaris_234846.tar I am calling the Ultrium drives with
Veritas as
/dev/rmt/0bn and /dev/rmt/1bn
The problem occurs when I attempt my second backup to the volume after the
media has
been ejected. The error I am seeing is:
/opt/openv/netbackup/logs/bptm/log.n
Subject: Backup on hostname - 86 started
Backup on client clientname by root failed.
Status = media position error.
I also get in /var/adm/messages
Jul 3 14:01:30 hostname unix: /pci@1f,2000/scsi@1 (glm2):
Jul 3 14:01:30 hostname Cmd (0xb3fb08) dump for Target 2 Lun 0:
Jul 3 14:01:30 hostname unix: /pci@1f,2000/scsi@1 (glm2):
Jul 3 14:01:30 hostname cdb=[ 0x17 0x0 0x0 0x0 0x0 0x0 ]
Jul 3 14:01:30 hostname unix: /pci@1f,2000/scsi@1 (glm2):
Jul 3 14:01:30 hostname pkt_flags=0x10000 pkt_statistics=0x60
pkt_state=0x7
Jul 3 14:01:30 hostname unix: /pci@1f,2000/scsi@1 (glm2):
Jul 3 14:01:30 hostname pkt_scbp=0x0 cmd_flags=0x60
Jul 3 14:01:30 hostname unix: WARNING: /pci@1f,2000/scsi@1 (glm2):
Jul 3 14:01:30 hostname Connected command timeout for Target 2.0
Jul 3 14:01:30 hostname unix: WARNING: ID[SUNWpd.glm.cmd_timeout.6017]
Jul 3 14:01:30 hostname unix: WARNING: /pci@1f,2000/scsi@1/st@2,0 (st16):
Jul 3 14:01:30 hostname SCSI transport failed: reason 'timeout': giving up
Jul 3 14:01:33 hostname unix: WARNING: /pci@1f,2000/scsi@1 (glm2):
Jul 3 14:01:33 hostname invalid reselection (2.0)
This only occurs with Veritas. When I use following unix tape commands
in sequence, such as:
tar -cvf /dev/rmt/1bn /opt
mt -f /dev/rmt/1bn rewind
mt -f /dev/rmt/1bn fsf 1
tar -cvf /dev/rmt/1bn /opt
mt -f /dev/rmt/1bn bsf 2
mt -f /dev/rmt/1bn fsf 2
time tar -cvf /dev/rmt/1bn /usr/test
etc. the drives work normally.
I have also made some changes to the st.conf file but the final cut
is:
"HP Ultrium", "HP Ultrium", "ULTRIUM";
ULTRIUM = 1,0x36,0,0xd639,4,0x00,0x00,0x00,0x40,3;
Veritas is researching the issue as is HP but have not offered any
fixes yet. I am assuming there are at least a few sites running this, or
a similar configuration. It appears that there is a problem when the
tape is remounted and Veritas attempts to read the header information
from the volume. The first time the tape is used, the backup proceeds
normally. If I attempt to use the same volume again, the attempt will
fail with a media position error and fail. It will then down the drive.
Thanks for your attention.
David Day
david.day AT sous DOT com
|