Tape Device Problems

[bjdobs]

Please Help ...

I am looking after a UNIX box from the eighties that is running 24/7 ... it has a lone-tar backup system to scsi tape ... a month ago it started to act up with backups failing every night (lights on front of drive flashing) so we rebooted the machine which cleared the flashing lights and it appeared the backups started working again however I just took a look at the backup log and it hasn't been working at all.

I did a ps -ef | grep ltar and for every night there was a process still pending since the reboot

I did a kill -9 for each of the processes found and then attemped to do a tape status ... this caused my session to hang

I started a new session and found there was also a lone-tar process hanging for each night so I did a kill -9 on these as well ... I then attempted to do another tape status and again the session hung

I now have three processes that appear to be asleep (kill -9, 2 or 3 will not clear them) one lone-tar and two tape statuses

Is there a way to kill these other then doing a reboot?
tapecntl -r is not doing anything (tape device reset)

Does sco have a generic device reset enable disable?

At this point I still don't know what is wrong with the backup process after running into one road block after another.

 

[edfair]

Try a kill -15.

But you will also need to do a reboot at some time shortly to clear stuff.

And take a look at crontabs -l to see what is called.

The flashing lights are generally error codes on the tape drive. You might want to start looking for a backup drive.
Ed Fair
unixstuff@juno.com
Any advice I give is my best judgement based on my interpretation of the facts you supply. Help increase my knowledge by providing some feedback, good or bad, on any advice I have given.

 

[bjdobs]

This is a live point of sale system 24/7 ... won't a kill -15 disrupt the entire system??? I will have to schedule an outage to reboot and if as you state a reboot is required anyway I guess I will just have to wait.

Thanx

 

[edfair]

Kill -15 is a sure kill. Doesn't touch the system. Unless you do it to the POS app.
from root "man kill". Ed Fair
unixstuff@juno.com
Any advice I give is my best judgement based on my interpretation of the facts you supply. Help increase my knowledge by providing some feedback, good or bad, on any advice I have given.

 

[bjdobs]

Kill -15 didn't work either ... according to the man page the kill command cannot kill any process that is sleeping due to a missed IRQ or waiting for an event ... in this case the processes have an S status which indicates they are asleep ... so I guess I have to schedule an outage to reboot the system and hope the tape drive hasn't packed it in.

 

[stanhubble]

Make sure you do a power off when you do the reboot.
It does sound like your tape drive has packed it in.

 

[bjdobs]

We have rebooted and hopefully gotten to the bottom of the problem ... in the /log directory is a number of files that are created by the Lone-tar process ... the ltar.log is only a summary of the events ... when I looked at the detail logs the real problem was revealed ... over the last two months the full backup went over the one tape limit and the process was asleep waiting for tape two ... because the ltar.M.cron script was not setup to pipe the request for second tape through to the consol this tidbit got lost in the detail logs ... this also explains why all the processes were still running over the last month as they were all waiting for the previous job to complete ... and this also explains why the kill command didn't work as I had killed the parent processes first ... when I kill the child process first the kill command works properly ... the only mystery now is why the tape drive lights would come on several weeks after the first process hung.

We have now changed to a selective backup until the tape system is replaced with a network wide DVD backup process (4.7G) which is under developement.

You wouldn't know off hand if the Lone-tar process can just create a backup file? If not I will just use cpio and compress.

Thanx