SUMMARY: C1533A & ufsdump on Solaris 2.3 ( SS20)

From: Administradores Sol (root@sol.info.unlp.edu.ar)
Date: Thu Oct 05 1995 - 11:22:40 CDT


Hello Managers:

                I solved my problem, thanks a lot to all who answered.

                Here is the original question:

["

                I'm having troubles with my system since I installed a C1533A ( HP Jet/Sure Store 6000 ). My config is the following:

Sun 1.05GB SCSI Disk internal. target 3.
Seagate Barracuda 4.2 Gb External. target 2
Sun CDROM internal target 6

The system was working perfectly, until I installed the DAT. The dat is external, daisy chained to the Barracuda. The DAT scsi number is 4.
I reconfigured the system by adding the following lines to /kernel/drv/st.conf, and doing a boot -r

tape-config-list =
        "HP C1533A","HP DDS2 4mm DAT", "HP-data2";
HP-data2 = 1,0x34,0,0x19,4,0x0,0x3,0x3,0x3,3;

The I rebooted and I got this message:

Sep 18 10:51:12 sol unix: WARNING: /iommu@f,e0000000/sbus@f,e0001000/espdma@f,400000/esp@f,800000/sd@3,0 (sd3):
Sep 18 10:51:12 sol unix: Error for command 'write' Error Level: Retryable
[..]
Sep 18 10:51:13 sol unix: Target 3.0 reducing sync. transfer rate
Sep 18 10:51:13 sol unix: polled command timeout
[..]
Sep 18 10:51:13 sol unix: WARNING: /iommu@f,e0000000/sbus@f,e0001000/espdma@f,400000/esp@f,800000/sd@3,0 (sd3):
Sep 18 10:51:13 sol unix: SCSI transport failed: reason 'timeout': retrying command

The strange thing is that I got errors from the internal disk!

scsiinfo gives me this:

esp0: sd2,0 tgt 2 lun 0:
        Synchronous(10.0MB/sec) Clean CanReconnect
esp0: sd3,0 tgt 3 lun 0:
        Synchronous(5.715MB/sec) Clean CanReconnect
esp0: sd6,0 tgt 6 lun 0:
        Synchronous(4.445MB/sec) Clean CanReconnect
esp0: st4,0 tgt 4 lun 0:
        Synchronous(10.0MB/sec) Clean CanReconnect

"]

---------------------------------------------------------------------------

It seems we have problems with the cable. We replaced the cable with a new and shorter one and now the system is working smoothly.

Now scsiinfo gives me:

esp0: sd2,0 tgt 2 lun 0:
        Synchronous(10.0MB/sec) Clean CanReconnect
esp0: sd3,0 tgt 3 lun 0:
        Synchronous(10.0MB/sec) Clean CanReconnect
esp0: sd6,0 tgt 6 lun 0:
        Synchronous(4.445MB/sec) Clean CanReconnect
esp0: st4,0 tgt 4 lun 0:
        Synchronous(10.0MB/sec) Clean CanReconnect

Here are the answers I got:

bismark@alta.jpl.nasa.gov (Bismark Espinoza)
steve@seaspace.com (Steve Clemons)

I got a long answer from misawa@physics.Berkeley.EDU (Shigeki Misawa)

Here it is ( edited )

-----

First, I don't think your tape-config-list entry is correct.
The white spaces in the first string ARE SIGNIFICANT. They must match
what the tape drive reports. The entry I have is :

           "HP C1553A", "HP DDS2 4mm DATloader", "HP-data2";

Second, you probably want to change the driver to handle longer
timeouts. The fourth number in the following line should be changed :

   HP-data2 = 1,0x34,0,0x19,4,0x0,0x3,0x3,0x3,3;

You might try one of the following (I don't what to bother with figuring
out which bit increase the time out. If you want to find out the exact
bit check the man pages and the approprate header file. :

HP-data2 = 1,0x34,0,0x1639,4,0x0,0x3,0x3,0x3,3;

or

HP-data2 = 1,0x34,0,0x9679,4,0x0,0x3,0x3,0x3,3;

(Oops, if you are using Solaris 2.3 then the third number needs to be 1024)

[...]

----

Regarding the configuration parameters, I have read this message on comp.unix.solaris:

Article: comp.unix.solaris.55873 Message-ID: <43r6kp$qsd@segfault.monkeys.com> From: rfg@monkeys.com (Ronald F. Guilmette) Subject: IMPORTANT! (was: Re: HP C1533 Parameters for Solaris 2.4 st.conf needed) Score: 100 First 60 lines: In article <SINDER.95Sep12204223@harpo.rrz.uni-koeln.de>, Andreas Sindermann <sinder@harpo.rrz.uni-koeln.de> wrote: >We have the following entry in /kernel/drv/st.conf: > >tape-config-list= > "HP C1533A", "HPC1533a", "hpc1533a"; >hpc1533a =1,0x34,1024,0x239,1,0x00,1; ^^^^ No! NO! NO! That third parameter should be a 0 for Solaris 2.4! Otherwise, no matter what you do, you will only get 1024 bytes per block on the tape... and that produces very slow performance when reading tapes which have been written with that (small) block size and probably... although I'm less sure of this... truly awful utilization of the tape. Note that that third parameter _had to be_ some non-zero value under Solaris ->2.3<- because of a kernel bug. But that kernel bug has been fixed in 2.4, and now you really need to set that third parameter to zero. If you don't believe me, just try _searching_ some of the backup tapes you have been making on Solaris 2.4 with 1024 in that third parameter position and see how fast the search progresses. You will be appalled! Then try putting a zero there, reboot, and try writing some new tapes with a _good sized_ blocking factor... like say for instance 48KB per block. (Don't even bother trying to use using 64KB per block or above. It won't work right.) If you are using tar or cpio, you can specify the blocking factor on the command line. Note however that under Solaris 2.4, if you have anything other than a zero in that third parameter position, whatever block size you tell tar or cpio to use will be ineffective, and the Solaris st driver will quietly breakup all blocks so that they come out to be exactly 1024 bytes long when they are actually written out to the tape. >It makes about 900KB/sec... Yea. That's what I thought too... and the drive _seems_ to run real fast as long as you are only _writing_ stuff. But try _reading_ some of your backup tapes that you have (perhaps unwittingly) made with only 1024 bytes per block. Can you say ``Slows to a crawl''? --

***************************************************************************

I posted another question, the first question how to rewind the tape. I was using mt eom wich, when used, went to the end of the media and the REWINDED the tape, because the default device is /dev/rmt/0. So now I'm using mt -f /dev/rmt/0n eom and it's working ok.

Thanks to these helpful people for answering:

Chris Royle <Chris.Royle@elmail.co.uk> sweh@mpn.com (Stephen Harris) "B C Hamshere" <B.C.Hamshere@newcastle.ac.uk> wezel@bio.vu.nl (Jos van Wezel) Eric Shafto <eshafto@caxy.lfa.lfc.edu> Mike Clarke <jmc@rripgpar.demon.co.uk> Scott Turvey <scottt@nacm.com> Jay Lessert <jayl@lattice.com> md@san-jose-1.firstimage.com (Joe Doyle) bismark@alta.jpl.nasa.gov (Bismark Espinoza) Ray Brownrigg <Ray.Brownrigg@isor.vuw.ac.nz>

The second question was about the transfer rate. I was getting about 3 hours to download 200mb. With the following configuration I get about 32 Mbytes in a minute.

from /kernel/drv/st.conf

tape-config-list = "HP C1533A","HP DDS2 4mm DAT", "HP-data2"; HP-data2 = 1,0x34,1024,0x1639,4,0x0,0x3,0x3,0x3,3;

Thanks a lot. Regards, Ramiro

Gustavo Guaragna - Ramiro Gonzalez Maciel | 'Suicide is painless, LIFIA - UNLP | it brings on many changes' C.C. 11 | MASH 4077th La Plata - Argentina | Te: +54 21 22 8252 Fax: +54 21 25 6809 | 'Ahhh, burritos?'



This archive was generated by hypermail 2.1.2 : Fri Sep 28 2001 - 23:10:34 CDT