Resetting SCSI

[Javier Ortiz Guajardo]

i receive this message in system log every 30 seconds aprox.

vmunix: SCSI: Resetting SCSI -- lbolt: 153930061, bus: 4

vmunix: SCSI: Reset detected -- lbolt: 153930061, bus: 4

how can i now what disks are connected to the bus: 4?

Thanks

[Kent Ostby]

You can usually see this by looking at ioscan -fn output.

It will show you which bus is bus 4.

Check the SCSI connections on that bus and also check the level of your SCSI patches.

Depending on your OS, there should be some patches that address SCSI resets.

[A. Clay Stephenson]

Look for the "device" data. You will see a long hex number. The first 2 hex digits represent the major device number. You can use lsdev to translate these (after converting the hex digits to decimal). The next two hex digit indicate the controller instance number, the next hex digit is the SCSI ID, the next hex digit is the LUN. The remaining digits are driver dependent.

[A. Clay Stephenson]

By the way, the 'lbolt' numbers simply is a timestamp of sort, the number of ticks since boot.

[Javier Ortiz Guajardo]

is this a problem with the SCSI controller or the disks?

[A. Clay Stephenson]

Insufficient data. The device number would tell a lot.

[Steven E. Protter · Accepted solution]

It could be the drive, controller or no some servers a device called a drive cage.

If the drive was powered off or unavailable longer than the timeout, lbolt

Of there was a problem with the controller that lasted longer than a timeout lbolt

A drive cage is a device on certain servers through which power and bandwidth go to the disks. I've had to replace that once, though the system blew a few disks before we nailed it down.

AS A. Clay said, insufficient data, if replacing a disk makes it stop for a year, its probably the the disk.

SEP

[A. Clay Stephenson]

Post a generous portion of the syslog. Enough, for example, to see if the errors are limited to one device. The problem is that the portion of the syslog that you thought was important is arguably the least significant. It could be something as simple as increasing the timeout on a PV.

[Javier Ortiz Guajardo]

thanks for your help. The problems was a damage disk (it was for mirror), i remove the lvols mirrored in this disk and i receive the message:

lvreduce -m 0 /dev/vg01/lvol12 /dev/dsk/c4t4d0
Logical volume "/dev/vg01/lvol12" has been successfully reduced.
vgcfgbackup: Unable to read the physical volume.: No such device or address
vgcfgbackup: Invalid LVMREC on Physical Volume /dev/rdsk/c4t4d0

then i remove the disk from VG.

Thanks.