Seen it many times,
Notice where your messages say

> unix: WARNING: /pci@1f,0/pci@1/scsi@1,1 (glm1):
> unix:  Connected command timeout for Target 1.0
> unix: WARNING: ID[SUNWpd.glm.cmd_timeout.6017]
> unix: WARNING: /pci@1f,0/pci@1/scsi@1,1 (glm1):
> unix:  Target 1 reducing sync. transfer rate
> unix: WARNING: ID[SUNWpd.glm.sync_wide_backoff.6014]
> unix: WARNING: /pci@1f,0/pci@1/scsi@1,1/sd@1,0 (sd17):

This is basically telling you that you have a slower device on the
scsi bus than the initiator can handle. 

A wild guess from your messages

> unix: WARNING: /pci@1f,0/pci@1/scsi@1,1/

This controller is a dual connect ultra wide (SE or DIFF) and
the hard disk drive [sd@1,0 (sd17)] is NOT an ultra wide device.

No worries though it does work just at a reduced scsi bus transfer rate.

The problem being that when the ENTIRE bus resets the NetBackup device
control daemon is told that the tape drive has a failed with a transport
failure or fatal timeout. Downs the drive. The problem may seem random 
because NB is not always trying to communicate with the tape drive when 
the bus resets.

There is a few things you can try.
1) Move your disk drive to the physical end of the scsi chain and if it
is a SE bus use a line forced perfect terminator. 

2) put your 4700 on its own bus. It looks like you have one available.
/pci@1f,0/pci@1/scsi@1

3)If the resets happen quite regular than replace the disk target 1 / sd17

A probable cause for the late collisions would be a switch hard set and
a Sun box using the auto negotiate feature on hme0.

Make sure that both the switch and U10 are hard set or auto negotiating.
If you would like the hme settings email me.

Good luck,
RF


Rasana Atreya wrote:
> 
> Hi,
> 
> I have a DLT4700 Mini Library connected to an Ultra 10 which is the NB
> server. OS is Solaris 2.5.1, NB version 3.2.
> 
> For some reason Netbackup keeps downing this drive randomly. When I use
> drive control to bring it back up, backups will proceed without problems.
> Also when the system is rebooted, almost always, the drive is downed.
> 
> Here's part of my syslog (I'm looking into the late collisions problem too):
> 
> unix: glm1:    Cmd (0x607bdd60) dump for Target 1 Lun 0:
> unix: glm1:            cdb=[ 0xa 0xc 0xb5 0x20 0xa 0x0 ]
> unix: glm1:    pkt_flags=0x4000 pkt_statistics=0x61 pkt_state=0x7
> unix: glm1:    pkt_scbp=0x0 cmd_flags=0x18e0
> unix: WARNING: /pci@1f,0/pci@1/scsi@1,1 (glm1):
> unix:  Connected command timeout for Target 1.0
> unix: WARNING: ID[SUNWpd.glm.cmd_timeout.6017]
> unix: WARNING: /pci@1f,0/pci@1/scsi@1,1 (glm1):
> unix:  Target 1 reducing sync. transfer rate
> unix: WARNING: ID[SUNWpd.glm.sync_wide_backoff.6014]
> unix: WARNING: /pci@1f,0/pci@1/scsi@1,1/sd@1,0 (sd17):
> unix:  SCSI transport failed: reason 'reset': retrying command
> unix: WARNING: /pci@1f,0/pci@1/scsi@1,1/sd@1,0 (sd17):
> unix:  SCSI transport failed: reason 'timeout': retrying command
> unix: SUNW,hme0: late collision
> unix: st12:    <Quantum DLT 4700 Mini Library>
> unix: st12 at glm1:
> unix:  target 5 lun 0
> unix: st12 is /pci@1f,0/pci@1/scsi@1,1/st@5,0
> unix: SUNW,hme0: late collision
> 
> I'd appreciate any pointers.
> 
> Regards,
> Rasana
> ______________________________________________________
> Get Your Private, Free Email at http://www.hotmail.com
> 
> _______________________________________________
> Veritas-bu maillist  -  Veritas-bu AT mailman.eng.auburn DOT edu
> http://mailman.eng.auburn.edu/mailman/listinfo/veritas-bu