Date: Sat, 14 Aug 1999 11:20:11 +0200 (CEST) From: Wilko Bulte <wilko@yedi.iaf.nl> To: karl@Denninger.Net (Karl Denninger) Cc: randy@psg.com, freebsd-scsi@FreeBSD.ORG Subject: Re: dump to dlt gets write error Message-ID: <199908140920.LAA53941@yedi.iaf.nl> In-Reply-To: <19990813191646.A57450@Denninger.Net> from Karl Denninger at "Aug 13, 1999 7:16:46 pm"
next in thread | previous in thread | raw e-mail | index | archive | help
As Karl Denninger wrote ...
> I've seen this kind of stupidity before and you're not going to like the
> problem or solution.
>
> Put the DLT on a different SCSI bus (different host adapter) from the disks.
>
> Specifically, separate the fast/wide and narrow SCSI devices.
>
> I've seen both DLTs and other "non-wide" devices have kittens with disks
> running fast/wide on the same SCSI bus. It usually manifests itself as
> an I/O error on the narrow device - which is exactly what you're getting.
I've been doing this for years and it works just fine:
FreeBSD 3.2-STABLE #5: Sun Aug 8 17:13:28 CEST 1999
root@yedi.iaf.nl:/usr/freebsd-stable-src/src/sys/compile/YEDI
[....]
da0: <DEC RZ1EF-CB (C) DEC 0371> Fixed Direct Access SCSI-2 device
da0: 40.000MB/s transfers (20.000MHz, offset 8, 16bit), Tagged Queueing Enabled
da0: 17365MB (35565080 512 byte sectors: 255H 63S/T 2213C)
evice to da0s2a
cd0 at ahc0 bus 0 target 4 lun 0
cd0: <TOSHIBA CD-ROM XM-5701TA 0557> Removable CD-ROM SCSI-2 device
cd0: 10.000MB/s transfers (10.000MHz, offset 8)
cd0: Attempt to query device size failed: NOT READY, Medium not present
da1 at ahc0 bus 0 target 1 lun 0
da1: <IBM DDRS-39130D DC1B> Fixed Direct Access SCSI-2 device
da1: 40.000MB/s transfers (20.000MHz, offset 8, 16bit), Tagged Queueing Enabled
da1: 8715MB (17850000 512 byte sectors: 255H 63S/T 1111C)
cd1: <PHILIPS CDD3600 CD-R/RW 2.00> Removable CD-ROM SCSI-2 device
cd1: 10.000MB/s transfers (10.000MHz, offset 15)
cd1: Attempt to query device size failed: NOT READY, Medium not present
sa3 at ahc0 bus 0 target 2 lun 0
sa3: <ARCHIVE 4326XX 27871-XXX 0322> Removable Sequential Access SCSI-2 device
sa3: 5.000MB/s transfers (5.000MHz, offset 15)
sa2 at ahc0 bus 0 target 5 lun 0
sa2: <DEC TZ88 (C) DEC D473> Removable Sequential Access SCSI-2 device
sa2: 10.000MB/s transfers (10.000MHz, offset 15)
sa0 at ahc0 bus 0 target 6 lun 0
sa0: <TANDBERG TDC 4200 00A1> Removable Sequential Access SCSI-2 device
sa0: 3.300MB/s transfers
The TZ88 is a DLT4000 btw, I also used a TZ87 which is a DLT2000. All
my tapes are in a Storageworks shelf.
> My guess is that the hardware on the narrow (and not-so-fast) device gets
> mightily confused by the shorter signal times (even though they're not
> aimed at that target) and randomly "freaks out" enough to botch an
> operation.
A DLT4000 is also a fast scsi device. My DLT2000 which is 5 Mb/sec also
worked just fine.
If I had to guess this is bad interconnect of some kind, or lousy
termination.
> Do you get any kind of DMESG log when the write *fails* (check it) or a
> console log of the actual error?
You can also pull the error logs from within the DLT drive itself. Try
the script below:
#!/bin/sh
# dltinfo: get more information out of your DLT tape drive.
#
# (C) 1996, Wilko Bulte, wilko@freebsd.org
#
# Warning: This script has only been tested on a DEC TZ87 & TZ88 DLT
#
# You need the DLT drive's OEM manual (or similar) to make
# sense out of some of the data reported.
# Please send any constructive comments by email to wilko@freebsd.org
Unit=2
## camcontrol(8) setup
Verbose="-v"
TimeOut="-t 3"
get_write_error_log() {
RetVal=`camcontrol cmd -n sa -u $Unit \
$Verbose \
$Timeout \
-c "4d 0 42 0 0 0 0 0 3f 0" \
-i 63 \
"{skip} *i4 \
{skip} *i4 \
{Corrected errors without substantial delay} i4 \
{skip} *i4 \
{Corrected errors with possible delay } i4 \
{skip} *i4 \
{Total errors } i4 \
{skip} *i4 \
{Total errors corrected } i4 \
{skip} *i4 \
{Total times correction algorithm processed} i4 \
{skip} *i4 \
{Total bytes processed } i8 \
{skip} *i4 \
{Total uncorrected errors } i4"
`
set $RetVal
echo "--- write errors ---"
printf "Corrected errors without substantial delay = %d\n" $1
printf "Corrected errors with possible delay = %d\n" $2
printf "Total errors = %d\n" $3
printf "Total errors corrected = %d\n" $4
printf "Total times correction algorithm processed = %d\n" $5
printf "Total bytes processed = %d\n" $6
printf "Total uncorrected errors = %d\n" $7
}
get_read_error_log() {
RetVal=`camcontrol cmd -n sa -u $Unit \
$Verbose \
$Timeout \
-c "4d 0 43 0 0 0 0 0 3f 0" \
-i 63 \
"{skip} *i4 \
{skip} *i4 \
{Corrected errors without substantial delay} i4 \
{skip} *i4 \
{Corrected errors with possible delay } i4 \
{skip} *i4 \
{Total errors } i4 \
{skip} *i4 \
{Total errors corrected } i4 \
{skip} *i4 \
{Total times correction algorithm processed} i4 \
{skip} *i4 \
{Total bytes processed } i8 \
{skip} *i4 \
{Total uncorrected errors } i4"
`
set $RetVal
echo "--- read errors ---"
printf "Corrected errors without substantial delay = %d\n" $1
printf "Corrected errors with possible delay = %d\n" $2
printf "Total errors = %d\n" $3
printf "Total errors corrected = %d\n" $4
printf "Total times correction algorithm processed = %d\n" $5
printf "Total bytes processed = %d\n" $6
printf "Total uncorrected errors = %d\n" $7
}
get_compression_log() {
# Assumption: from the results observed in testing it lookse
# like the residual counts are in kBytes (and not
# in Mbytes as the TZ87 manual tells us).
RetVal=`camcontrol cmd -n sa -u $Unit \
$Verbose \
$Timeout \
-c "4d 0 72 0 0 0 0 0 4c 0" \
-i 76 \
"{skip} *i4 \
{skip } *i4 \
{Read compression ratio (* 100 %) } i2 \
{skip } *i4 \
{Write compression ratio (* 100 %) } i2 \
{skip } *i4 \
{Total host Mbytes reads } i4 \
{skip } *i4 \
{Total host kbytes read residual } i4 \
{skip } *i4 \
{On tape Mbytes read } i4 \
{skip} *i4 \
{On tape kbytes read residual } i4 \
{skip} *i4 \
{Host requested Mbytes written } i4 \
{skip} *i4 \
{Host requested kbytes written residual } i4 \
{skip} *i4 \
{On tape Mbytes written } i4 \
{skip} *i4 \
{On tape kbytes written residual } i4 "
`
set $RetVal
echo "--- compression statistics ---"
printf "Read compression ratio = %d %%\n" $1
printf "Write compression ratio = %d %%\n" $2
printf "Total host Mbytes read = %d\n" $3
printf "Total host kbytes read residual = %d\n" $4
printf "On tape Mbytes read = %d\n" $5
printf "On tape kbytes read residual = %d\n" $6
printf "Host requested Mbytes written = %d\n" $7
printf "Host requested kbytes written residual = %d\n" $8
printf "On tape Mbytes written = %d\n" $9
printf "On tape kbytes written residual = %d\n" $10
}
get_read_error_log
echo
get_write_error_log
echo
get_compression_log
echo
It is a quick hack but it works for me. Like in:
Mon Aug 9 00:32:07 CEST 1999
DUMP: Date of this level 0 dump: Mon Aug 9 00:32:07 1999
DUMP: Date of last level 0 dump: the epoch
DUMP: Dumping /dev/rda1c (/local2) to /dev/nrsa2
DUMP: mapping (Pass I) [regular files]
DUMP: mapping (Pass II) [directories]
DUMP: estimated 2685410 tape blocks.
DUMP: dumping (Pass III) [directories]
DUMP: dumping (Pass IV) [regular files]
DUMP: 12.18% done, finished in 0:36
DUMP: 27.84% done, finished in 0:25
DUMP: 43.49% done, finished in 0:19
DUMP: 59.02% done, finished in 0:13
DUMP: 73.04% done, finished in 0:09
DUMP: 86.80% done, finished in 0:04
DUMP: 99.85% done, finished in 0:00
DUMP: DUMP: 2686684 tape blocks on 1 volumes(s)
DUMP: finished in 2104 seconds, throughput 1276 KBytes/sec
DUMP: level 0 dump on Mon Aug 9 00:32:07 1999
DUMP: Closing /dev/nrsa2
DUMP: DUMP IS DONE
Mon Aug 9 01:07:41 CEST 1999
--- read errors ---
Corrected errors without substantial delay = 0
Corrected errors with possible delay = 0
Total errors = 0
Total errors corrected = 0
Total times correction algorithm processed = 0
Total bytes processed = 0
Total uncorrected errors = 0
--- write errors ---
Corrected errors without substantial delay = 0
Corrected errors with possible delay = 0
Total errors = 105
Total errors corrected = 105
Total times correction algorithm processed = 0
Total bytes processed = 0
Total uncorrected errors = 0
--- compression statistics ---
Read compression ratio = 0 %
Write compression ratio = 100 %
Total host Mbytes read = 0
Total host kbytes read residual = 0
On tape Mbytes read = 0
On tape kbytes read residual = 0
Host requested Mbytes written = 9196
Host requested kbytes written residual = 196608
On tape Mbytes written = 9196
On tape kbytes written residual = 0
The most interesting part is the Total errors thingy. I've seen that
sky rocket with bad media or DLT drives with a bad head.
IMHO this kind of errorlogging would be cool to have in any standard
shape. Like VMS does, or in a quite different form, DEC Unix, eh Tru64.
Really useful in case you have hardware problems.
> Karl Denninger (karl@denninger.net) Web: childrens-justice.org
>
>
> On Fri, Aug 13, 1999 at 05:05:17PM -0700, Randy Bush wrote:
> > asus p2b-ds 2x350mhz, 128mb
> > two barracudas
> > quantum dlt2000
> > 4.0-currnt of 99.04.03
> >
> > rip.psg.com:/# /do-dump
> > ...
> > DUMP: Date of this level 0 dump: Fri Aug 13 16:07:11 1999
> > DUMP: Date of last level 0 dump: the epoch
> > DUMP: Dumping /dev/rccd5c (/usr) to /dev/nrsa0
> > DUMP: mapping (Pass I) [regular files]
> > DUMP: mapping (Pass II) [directories]
> > DUMP: estimated 2951716 tape blocks.
> > DUMP: dumping (Pass III) [directories]
> > DUMP: dumping (Pass IV) [regular files]
> > DUMP: 11.37% done, finished in 0:38
> > DUMP: 25.73% done, finished in 0:28
> > DUMP: 39.41% done, finished in 0:23
> > DUMP: 51.77% done, finished in 0:18
> > DUMP: 64.66% done, finished in 0:13
> > DUMP: 76.19% done, finished in 0:09
> > DUMP: 88.92% done, finished in 0:04
> > DUMP: write error 2700020 blocks into volume 1
> > DUMP: Do you want to restart?: ("yes" or "no")
> >
> > usually a LOT more fits on a tape, like four machines more.
> >
> > i ran the cleaning tape. i tried different tapes from different batches,
Don't ever run cleaning tapes on a DLT drive unless the 'Use cleaning tape'
LED comes on. Cleaning tapes are really bad news for the DLT heads if they
are run on a regular basis. They consist of a more or less normal data tape
that did not get it's final polishing steps in manufacturing. They are
quite abrasive and do bad things to non-dirty heads.
> > including one that worked in the past. it breaks at different places, but
> > always much of the way through that partition.
> >
> > clues solicited.
> >
> > randy
> >
--
| / o / / _ Arnhem, The Netherlands - Powered by FreeBSD -
|/|/ / / /( (_) Bulte WWW : http://www.tcja.nl http://www.freebsd.org
To Unsubscribe: send mail to majordomo@FreeBSD.org
with "unsubscribe freebsd-scsi" in the body of the message
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?199908140920.LAA53941>