Topic: Help in getting flexbackup going...

[Rob Hillis]

I've recently bought myself a secondhand WangDAT 3400 tape drive.  I've verified that it works under Win2k, but I can't for the life of me get it going under e-smith.  Flexbackup runs as follows:-

-----8><------
[root@e-smith /root]# flexbackup -fs -all

flexbackup version 0.9.8
/etc/flexbackup.conf syntax OK

|------------------------------------------------
| Trying "mt defblksize" instead of "mt setblk"
Error setting block size
Neither of these commands worked:
  mt -f /dev/nst0 setblk 0
  mt -f /dev/nst0 defblksize 0
-----8><------

I'm using a Tekram DC-315U to drive it, and although it's difficult to ensure I'm seeing everything, during bootup, the kernel *does* report the existance of a SCSI card.  The tape drive is set as SCSI device 7, and is the only SCSI device in the system.

Does anyone have any ideas?

[Glenn]

I'm not familiar with your SCSI board, but if it is of the older 8 channel narrow SCSI type you should not use SCSI ID#7 as this is reserved for the controller itself. Try switching the tape drive's ID to 3 or 4 for example and try again. Hope this works for you.


Glenn

[Rob Hillis]

Glenn wrote:

> I'm not familiar with your SCSI board, but if it is of the
> older 8 channel narrow SCSI type you should not use SCSI ID#7
> as this is reserved for the controller itself. Try switching
> the tape drive's ID to 3 or 4 for example and try again. Hope
> this works for you.

It's a fairly new SCSI card, and I'm actually wondering whether e-smith is detecting it properly at all, now.  Linux-Mandrake has no problems at all with it, but RedHat 7 doesn't detect it...

I've never yet successfully recompiled a kernel, but I guess it's time to learn...

[Michael Doerner]

Rob,

>I've never yet successfully recompiled a kernel, but I guess it's time to learn...
I doubt that you will have to go that far ...

I would start to do more tests with the mt (Magnetic Tape) command i.e. 'mt -f /dev/nst0 status'. Have a look at the mt command options with 'man mt' and see whether the device reacts on a rewind operation for example?

If that works, it might be useful to have a look in /etc/flexbackup.conf. There are definitions for blocksize and variable-blocksize.
I had to play with these blocksizes when I tested different DAT devices quite a while ago.

Hope that helps a bit.

Regards,
Michael Doerner

P.S.: That's the output of a 'mt -f /dev/nst0 status here (it's an older DAT model, a WangDAT 3200):
[root@tga1 /root]# mt -f /dev/nst0 status
SCSI 2 tape drive:
File number=0, block number=0, partition=0.
Tape block size 0 bytes. Density code 0x13 (DDS (61000 bpi)).
Soft error count since last status=0
General status bits on (41010000):
 BOT ONLINE IM_REP_EN

[Rob Hillis]

Michael Doerner wrote:
>
> >I've never yet successfully recompiled a kernel, but I guess
> >it's time to learn...
> I doubt that you will have to go that far ...

Unfortunately it looks *very* much like that...

> I would start to do more tests with the mt (Magnetic Tape)
> command i.e. 'mt -f /dev/nst0 status'. Have a look at the mt
> command options with 'man mt' and see whether the device
> reacts on a rewind operation for example?

No... it claims that the device doesn't exist.  Having now watched the bootup sequence very closely, I can't see any mention of my DAT drive being detected, and I'm *far* from certain that the SCSI message I see during bootup is related to my Tekram card.

> P.S.: That's the output of a 'mt -f /dev/nst0 status here
> (it's an older DAT model, a WangDAT 3200):

Mine's a 3400, so there should be next to no difference...

What kind of SCSI card are you using?

[Michael Doerner]

I am also using a Tekram card (U2W) but can't remember the exact model and don't want to open the server at the moment.

As Glenn already recommended, I would definitely change the SCSI ID for your tape device because SCSI ID 7 is by default used from most SCSI controllers themselves (also from the Tekram's as you see below).

Here is an extract from my /var/log/dmesg file (the ncr53c8xx is the driver for the automatically detected SCSI controller):

ncr53c8xx: at PCI bus 0, device 9, function 0
ncr53c8xx: 53c895 detected
ncr53c895-0: rev=0x01, base=0xdf000000, io_port=0xd800, irq=5
ncr53c895-0: ID 7, Fast-40, Parity Checking
ncr53c895-0: on-chip RAM at 0xde800000
ncr53c895-0: restart (scsi reset).
ncr53c895-0: Downloading SCSI SCRIPTS.
scsi0 : ncr53c8xx - version 3.2a-2
scsi : 1 host.
  Vendor: QUANTUM   Model: ATLAS IV 9 WLS    Rev: 0B0B
  Type:   Direct-Access                      ANSI SCSI revision: 03
Detected scsi disk sda at scsi0, channel 0, id 1, lun 0
ncr53c895-0-<4,0>: quirks=2.
  Vendor: WangDAT   Model: Model 3200        Rev: 02.2
  Type:   Sequential-Access                  ANSI SCSI revision: 02
ncr53c895-0-<1,0>: tagged command queue depth set to 8
ncr53c895-0-<1,*>: FAST-40 WIDE SCSI 80.0 MB/s (25 ns, offset 31)
SCSI device sda: hdwr sector= 512 bytes. Sectors= 17942584 [8761 MB] [8.8 GB]

The controller was already detected and working in the older E-smith version 4.0 but at those days the driver supported only 40MB. Since Redhat 7.0, it's supporting 80MB.

Let us know about your progress. I still don't see that you will have to recompile the kernel
(BTW your controller might be a different model than my one. Have you checked Redhat's hardware compatibility for it?)

Regards,
Michael Doerner

[Rob Hillis]

> I am also using a Tekram card (U2W) but can't remember the exact model and
> don't want to open the server at the moment.

Awww...

> As Glenn already recommended, I would definitely change the SCSI ID for
your
> tape device because SCSI ID 7 is by default used from most SCSI
controllers
> themselves (also from the Tekram's as you see below).

That's been done, and yes, that was causing problems, even under Win2k and
Mandrake.

> Here is an extract from my /var/log/dmesg file (the ncr53c8xx is the
driver
> for the automatically detected SCSI controller):

Didn't know about that log file...  *very* useful.  Well that confirms
it...

scsi : 0 hosts.
scsi : detected total.

> Let us know about your progress. I still don't see that you will have to
> recompile the kernel
> (BTW your controller might be a different model than my one. Have you
> checked Redhat's hardware compatibility for it?)

It lists a DC-390F, but that's all.  Mine's a DC-315U.

The *really* annoying thing is that Mandrake (based on Redhat) finds it
during installation, and installs it without a hitch.  I've got a copy of
the module (dc395x_trm.o) which Mandrake installs, but can't get E-Smith to
use it...  modprobe comes up with:

[root@e-smith scsi]# modprobe dc395x_trm
/lib/modules/2.2.16-22/scsi/dc395x_trm.o: unresolved symbol scsi_register_R40b2f033
/lib/modules/2.2.16-22/scsi/dc395x_trm.o: unresolved symbol
scsi_unregister_Red9afb5b
/lib/modules/2.2.16-22/scsi/dc395x_trm.o: unresolved symbol
scsicam_bios_param_R7f410f04
/lib/modules/2.2.16-22/scsi/dc395x_trm.o: insmod
/lib/modules/2.2.16-22/scsi/dc395x_trm.o failed
/lib/modules/2.2.16-22/scsi/dc395x_trm.o: insmod dc395x_trm failed

I just don't know enough about integrating modules into the kernel...
The thing which really irritates me too is the damned box specifically says
drivers are available for RedHat.

[Michael Doerner]

Rob,

No, we don't give up here yet ....

Were the Mandrake and Win2k installations on the same box (hardware) where you are running E-smith now?

I am just thinking about a possible hardware/BIOS setup problem.
Could it be that there is no more IRQ left in the Bios to be assigne for the PCI slots or that particular PCI slot for the controller or something like that?

By starting the PC, it should tell you about the SCSI controller and it's attached devices before any OS is started.

Just some other thoughts....

[Rob Hillis]

Michael Doerner wrote:

> No, we don't give up here yet ....
> Were the Mandrake and Win2k installations on the same box
> (hardware) where you are running E-smith now?

Win2k: no. Mandrake: yes.

> I am just thinking about a possible hardware/BIOS setup
> problem.

Possible, but extremely unlikely.

> Could it be that there is no more IRQ left in the Bios to be
> assigne for the PCI slots or that particular PCI slot for the
> controller or something like that?

Once again, I doubt it..

> By starting the PC, it should tell you about the SCSI
> controller and it's attached devices before any OS is started.

Not if the controller doesn't have a BIOS in it...

[Michael Doerner]

>> By starting the PC, it should tell you about the SCSI
>> controller and it's attached devices before any OS is started.

>Not if the controller doesn't have a BIOS in it...

Correct.
Seems another one has to jump in now since I  don't know any further.

Regards,
Michael Doerner