tar: error is not recoverable: exiting now

[James Roberto]

Hi,

Operating System: RHEL AS4 U5
Kernel: 2.6.9-55.ELsmp
Aim: To backup data onto my LTO2 tape drive from file server over the network.
System setup: nfs
Size of Directory: 105GB
command: tar -cvf "device" "directory"
Tar Version: tar-1.14-12.RHEL4
Error message: /var/log/messages
- kernel: st0: Error with sense data: Current st0: sense key Medium Error
- kernel: Additional sense: Write error
- kernel: st0: Error on write filemark.
- kernel: Additional sense: Recorded entity not found

Can anyone help me please??
thanks
James

[Stuart Browne · Accepted solution]

Interesting.

Does this happen immediately, or after 'some time'?

What size tapes are you using in the drive?

Does this happen on all the tapes you've tried it with?

When did you last clean the drive?

This error says to me that the system can't write to the tape, so it exit's.

[James Roberto]

Thanks Stuart for replying.

It happens after 'some time'. It seems the tar is working, then after 5 mins or so it starts to give me the error message: input/output error.Then the Tar stops.

Im using 400GB HP LTO 2 tapes with a HP LTO2 tape drive.

Yes this happens on all the tapes I've tried it with.

I cleaned the tape drive before i started my tar backup. I often clean the tape drive every fortnight. Depends how i often i do my backups.

That too i also interpret that the system can't write to the tape.
But get this, i recently backuped my redhat iso's on the server. Ran the same command tar -cvf "device" "directory" and it works. The directory was 15GB......

I just dont quite understand why it won't do more than 15GB.

FYI My backups use to work. I use to backup 200GB files. But I think it may be something to do with NFS,NIS or maybe the network. any suggestions??

thanks James

[Stuart Browne]

Ok, small amounts which work is interesting. The next thing to try is a reboot, see if it comes back.

ALso check that the cables haven't been moved within the server, and that tne terminators on the SCSI chain are all good.

Curious question. Have you done a kernel upgrade since it last worked ?

What is the SCSI controller anyway?

[Ivan Ferreira]

I had similar problems with network backups once. The solution was to find the optimal block size for the device.

[James Roberto]

I've tried a reboot. Device is still there. Im actually doing the tar backup from a client workstation.

Terminators are all good. I thought it may have been the hardware.

Do you mean kernel upgrade on my server(where the 105GB data is), or client machine where im trying the backup.

SCSI controller?? i forgot the command to type in. Is it lsmod or modprobe. cant rememeber.


How do you set the block size for the device? Can you set it to unlimited. Is the command i use set blk "specify how big".

FYI
root@walcott]# mt -f /dev/nst0 stat
SCSI 2 tape drive:
File number=1, block number=0, partition=0.
Tape block size 0 bytes. Density code 0x42 (no translation).
Soft error count since last status=0
General status bits on (81010000):
EOF ONLINE IM_REP_EN

[Stuart Browne]

Client where you're doing the backup.

Block size is set to 0, which basically uses whatever the tape drive wants to use.

As you are pulling it across the network, it may pay to fix the block size to something smaller (I've used 5120 and 10240 in the past when going to an Ultrim drive), but the name of the game is expirmentation.

The fact that it worked in the past seems to lean away from a block size issue, but it's still worth investigating.

Now, to find out what device you're using, your friends are 'lsmod', 'lspci' and 'cat /proc/scsi/scsi'.

[James Roberto]

Ok... so i might change the block size to 10240. and see how that goes.

so far i just ran a tar backup job. I just dont know why it fails.

tar: /dev/nst0: Cannot write: Input/output error
tar: Error is not recoverable: exiting now
tar complete

So once i change the blocksize is there any way i can see how the network is affected or i can see why it fails besides using tial -f /var/log/messages

[Stuart Browne]

You'd need to set up some sort of monitoring to see the network throughput ('mrtg' is your friend), if you've not already got it somewhere.

There are some X tools which will do it for you on the fly, but I've never used them.

'dmsg' may give you some errors, but '/var/log/messages' will probably only tell you what you already know.

[Peter_505]

Do you have the Library and Tape Tools?

If you don't I'd download them and check the health of the drive.

http://h20000.www2.hp.com/bizsupport/TechSupport/SoftwareDescription.jsp?lang=en&cc=us&prodTypeId=12169&prodSeriesId=316040&prodNameId=1849276&swEnvOID=2025&swLang=8&mode=2&taskId=135&swItem=co-52769-1