What is meaning of bioerrors / scsi_bioerrors in a panic dump?

[Stan Sieler]

Hi,

Noticed during a q4 "run WH" analysis of an HP-UX 11.11 panic dump:

We have buffer I/O errors!!
bioerrors: 1188 nfsbioerrors: 0
scsi_bioerrors: 18624

I can't find any info on the above things,
but just intuitively they look menacing :)

Any info is appreciated.

thanks,

Stan
sieler@allegro.com

[Jino.P.V · Accepted solution]

This means is that the count of completed I/Os where B_ERROR was set.

What this means is that when an I/O was completed by a disk driver the driver
set B_ERROR in the I/O buffer. Since this value is only a count of the errors,
the errors could have happened at anytime since boot. The type of errors are
not identified. The errors could have been anything from from a media error to
a timeout because of activity or lack of a response from a device.

By itself this figure means very little. A very large value could possibly
indicate a disk, cable or controller problem. There may be other related
factors, for example a disk may have failed on the system and was replaced
online -> scsi_bioerror may have incremented to a high value but that value
may not be relevant to your current problem.

You would need to investigate if there are I/Os outstanding for a long time at
an LVM or SCSI level first (suggestion: use lvm.pl in q4) before deciding if
the value in scsi_bioerror is meaningful to the problem at hand.

q4's wsioscsi.pl can be used to show I/O at the SCSI level (this q4 perl script
is useful only for WSIO (e.g. HSC, EISA, PCI) buses, not for HP-PB (aka NIO)
buses).

The logtool utility in the diagnostics may be useful in determining of there
is a hardware problem on the system.

The Bioerror can come because the latest firmware is not installed.

Hope it is clear,

Jino

[Stan Sieler]

thanks, sounds like good info!