Topic: Possible Pickle - RAID and Tape drive

[Dave D]

Hi,

Firstly, thanks to all you  posters.  Informal user support is an excellent model.

Here is the situation.  I recently installed SME 5.6 (server only mode) and performed all the updates. Additionally, I have installed the Advanced flexbackup RPM's.

I decided to use software RAID - 2 identical drives, both on the MASTER channel of different IDE controllers.  I also have a Seagate STT20000A IDE tape drive installed on the SLAVE of the secondary IDE and a DVD-ROM on the SLAVE channel of the primary IDE.

I took a look at a recent log file for flexbackup and saw errors in the listing.

Here is the listing:

|------------------------------------------------
| Trying "mt defblksize" instead of "mt setblk"
| Doing level 0 backup of all using dump
| Retensioning tape...
| Rewinding & erasing tape...
| Creating index key 200306250203.34
| Tape #0
| Filesystems = / /boot
|------------------------------------------------
At block 21.
|------------------------------------------------
| File number 1, index key 200306250203.34
| Backup of: /
| Date of this level 0 backup: Wed Jun 25 02:04:06 2003
| Date of last level 0 backup: the epoch
|------------------------------------------------
| (dump -0 -b 10 -a -f - / | gzip -4) | buffer -m 3m -s 10k -u 100 -t -p 75 -B \
|  -o /dev/nst0
|------------------------------------------------
  DUMP: Date of this level 0 dump: Wed Jun 25 02:04:06 2003
  DUMP: Dumping /dev/md1 (/) to standard output
  DUMP: Added inode 8 to exclude list (journal inode)
  DUMP: Added inode 7 to exclude list (resize inode)
  DUMP: Label: /
  DUMP: mapping (Pass I) [regular files]
  DUMP: mapping (Pass II) [directories]
  DUMP: estimated 12583019 tape blocks.
  DUMP: Volume 1 started with block 1 at: Wed Jun 25 02:04:36 2003
  DUMP: dumping (Pass III) [directories]
  DUMP: dumping (Pass IV) [regular files]
  DUMP: 1.60% done at 670 kB/s, finished in 5:07
  DUMP: 3.60% done at 755 kB/s, finished in 4:27
  DUMP: 4.95% done at 692 kB/s, finished in 4:47
  DUMP: 6.50% done at 681 kB/s, finished in 4:47
  DUMP: 8.60% done at 721 kB/s, finished in 4:25
buffer (writer): write of data failed: Input/output error
bytes to write=10240, bytes written=-1, total written     681060K
  DUMP: Broken pipe
  DUMP: The ENTIRE dump is aborted.
|------------------------------------------------
| Backup start: Wed Jun 25 02:04:06 2003
| Backup end:   Wed Jun 25 02:30:00 2003
|------------------------------------------------
/dev/nst0: Input/output error
mt: The device is offline (not powered on, no tape ?).
|------------------------------------------------
| File number 2, index key 200306250203.34
| Backup of: /boot
| Date of this level 0 backup: Wed Jun 25 02:30:11 2003
| Date of last level 0 backup: the epoch
|------------------------------------------------
| (dump -0 -b 10 -a -f - /boot | gzip -4) | buffer -m 3m -s 10k -u 100 -t -p \
|  75 -B -o /dev/nst0
|------------------------------------------------
  DUMP: Date of this level 0 dump: Wed Jun 25 02:30:11 2003
  DUMP: Dumping /dev/md0 (/boot) to standard output
  DUMP: Added inode 8 to exclude list (journal inode)
  DUMP: Added inode 7 to exclude list (resize inode)
  DUMP: Label: /boot
  DUMP: mapping (Pass I) [regular files]
  DUMP: mapping (Pass II) [directories]
  DUMP: estimated 4675 tape blocks.
  DUMP: Volume 1 started with block 1 at: Wed Jun 25 02:30:11 2003
  DUMP: dumping (Pass III) [directories]
  DUMP: dumping (Pass IV) [regular files]
buffer (writer): write of data failed: Input/output error
bytes to write=10240, bytes written=-1, total written          0K
  DUMP: Broken pipe
  DUMP: The ENTIRE dump is aborted.
|------------------------------------------------
| Backup start: Wed Jun 25 02:30:11 2003
| Backup end:   Wed Jun 25 02:30:14 2003
|------------------------------------------------
/dev/nst0: Input/output error
mt: The device is offline (not powered on, no tape ?).
|------------------------------------------------
| Rewinding...
/dev/nst0: Input/output error
mt: The device is offline (not powered on, no tape ?).
|------------------------------------------------

Searching the forums indicated this has been seen before, and a potential solution is to run

/sbin/e-smith/db configuration setprop flexbackup
erase_rewind_only false

/sbin/e-smith/signal-event console-save

I just ran it and will see tomorrow if  it solves the errors.

I did a command line extract of the tape, and it looks complete for the 25th, but the error messages were there in the og as well....

So, I am wondering if it imperative that the Seagate IDE tape drive be installed on the Master of one of the IDE connectors in order to get reliable backups ?

If I do this- switch the master/slave for one HD and the tape drive , will I mess up my software RAID, and if I do mess it up, is there a way to re-establish the mirror without a major system outage ?  I seem to remember reading somewhere that the RAID arrays are auto-detected but my memory is really not all that good.

Any and all thoughts/comments are welcome.

Dave

[James Roberts]

Welcome to the  ATAPI backup problem. I've been struggling with this one this one for months now. There are many comments on the board but no solutions (but see below).

Re: ATAPI/IDE Seagate (or other) Travan as backup device - seems such a neat IDEa. doesn't it? Two HDD on Master/master, CD and tape on respective slaves. Here's the most recent comment I got from Seagate on this ( 14 June 03):

"We recommend that you never Slave this tape unit to a Hard Drive. If there
 is not another ATAPI/IDE device present, you can set the tape as Master on
 the Primary. If there is an HD present, the tape needs to be set as Master
 on the Secondary. If there is a device already on the secondary such as a
 CD-ROM, you need to Slave the tape drive to the CD-ROM. Make sure you have
 the port set to NONE or NOT INSTALLED in the Computers BIOS and not
 AUTOMATIC or USER."

This seems a fairly clear 'no' to me. Which means running the software raid on one controller which is of course a disaster speedwise. Hardware raid (or SCSI backup), welcome back!

The possible issues I have identified with Travan are as follows:

1. ide-scsi emulation (see support on Seagate site)
2. module loading timeout issues
3. various issues with the e-smith scripts
4. various issues with flexbackup itself.

The first is easily resolved. The second I may have a fix for, still testing but it's working for me. The third - there was a revised flexbackup script issued for 5.5 but installed as standard in 5.6 (which didn't fix my problem).

Finally, there is a brand new version of flexbackup (1.6) with mbuffer on CharlieBrady's download page but I want to know if it works on 5.5 and 5.6 before installing it.

OTOH with that error mesj you may just have a bad drive or media. Try locking down the issue more precisely.

JR

[Chris Woods]

Get another backup unit preferably SCSI.  I have 3 seagate Travan 40's that I'm currently using as paper weights and book ends they were replaced with Exabyte VXA-1 units and they work great.

[Kelvin]

Hi Dave,

Follow this thread :

http://forums.contribs.org/index.php?topic=17612.msg68690#msg68690

and test your backup again with tar instead of dump and see if you still get a problem.

Kelvin

[Charlie Brady]

Kelvin wrote:

> Follow this thread :
>
> http://forums.contribs.org/index.php?topic=17612.msg68690#msg68690
>
> and test your backup again with tar instead of dump and see
> if you still get a problem.

The dump problem (likely to be seen only on a very busy machine) would cause wrong data to be associated with a file. It wouldn't/couldn't cause an I/O error in writing to tape.

Charlie

[Kelvin]

Charlie wrote:

>It wouldn't/couldn't cause an I/O error in writing to tape

Yes, that's true, however, I have seen so far 2 occasions where a dump backup did not work while a tar one did.

Kelvin

[James Roberts]

Hi Charlie:

Does the new version of flexbackup (1.6) with mbuffer on on your page work on 5.5 and 5.6? If not I won't bother trying it.

Thanks!

JR

[James Roberts]

Hi Charlie:

Does the new version of flexbackup (1.6) with mbuffer on on your page work on 5.5 and 5.6? If not I won't bother trying it.

Thanks!

JR

[James Roberts]

Hi Chris:

> I have 3 seagate Travan 40's that I'm currently using as paper weights and book >ends

Great! If you are not using them you wouldn't like to send them to me for the testing I'm doing, perhaps?

We do have some tens of these units working, but in some circumstances they don't ('e-smith' circumstances, in fact - they work otherwise). We have also had problems with various SCSI devices in similar circumstances.

If you *can* do this then contact me off-list for postal address, etc. (but note I'm in Europe).

JR.

[Dave D]

Just to let you know that I haven't forgotten about this, here is a small update.

The attempted tar (thanks Kelvin) also failed with an I/O error. Hmmm,,,,
maybe the tape is bad. Though maybe using the drive as a slave is also causing it.....

Still waiting to test with new tape.  Should happen soon.  Server is in an environment that I don't have easy access to.

I will update this thread when I know more.

Backups of client data are fundamental as we all know...

[Kelvin]

Hi Dave,

>The attempted tar (thanks Kelvin) also failed with an I/O error.

Oh well, it was worth a try as I said, even though it did not fix your problem, it at least tells you that it is likely something to do with the drive itself or the tapes. Process of elimination .

By the way, James is right about not connecting the drives as slaves to the HDDs. This also applies to the CD-ROM. Not because this will cause backups to fail, but more so because the way IDE works, the entire IDE bus speed is determined by the slowest member of the chain (ie. your Tape drive and CD-ROM). If your are using software RAID, consider installing an additional IDE controller that is supported by the kernel and move the tape / cd-rom drives there or install the software raid on drives connected to the additional controller (during initial SME installation that is).

Kelvin