Hi managers,
Never too late !
The problem is not completely solved, but we have the
informations!
Thanks to Alan, Tom, Trevor and Christophe.
Here are their suggestions :
* When you get I/O errors, you should look at the error log
to see what the errors are. I've seen cases where there
weren't any logged device errors, but that is rare. If
using uerf(8) to format the error log, use the "-o full"
option to get the full device error information. On V4
and later prefer DECevent (the dia command) to format the
error log.
* I believe the lifespan of a 120M tape is 2000 passes which translates to
about 1000 reads and 1000 writes or a combination thereof. Usually if we
encounter error such as this we do a manual check of the filesystem and
destroy the tape replacing it with a new one. If the error persists I
would suggest a call in to Dec.
(2000 passes on a DLT drive does not translate to 1000 reads and 1000
writes since DLT technology takes several passes when manipulating
the tape)
* That's probably a write error on the tape media. Since you are
probably running the backups in a script, there is no controlling tty
for the vdump utility to talk to, so it gives up.
The tapex utility can be used to verify tape media. There is a man page
for it.
The TLZ09 works fine with vdump and with 120m tapes, but if you have a bad
spot on the tape, it's not going to just keep writing.
The life expectancy of the tape is specified by the vendor; you should get
that information from whoever is selling you the tapes (including Digital
or an authorized reseller). It can vary from one batch to another. But
it's a lot longer than you should need to worry about. Tapes generally
are either OK or not -- that is, if a tape is good, it usually won't "wear
out" unless you are re-using it constantly.
The original posting was :
> We backup different directories on our advfs volumes "utilisateurs"
> mounted on "/home" like that :
>
> /sbin/vdump -0uq -CDN -f /dev/nrmt0h /home/user1
> /sbin/vdump -0uq -CDN -f /dev/nrmt0h /home/user2
> ...
> /sbin/vdump -0uq -CDN -f /dev/nrmt0h /home/userN
>
> (we backup, at one time, only some users on /home, and backup
> other users on another time in the same day)
>
> The device is a TLZ09 (so, 4-8 Go) with 120 m tapes.
>
> Frequently we have problems like that :
>
> vdump: unable to write to device </dev/nrmt0h>; [5] I/O error
> vdump: unable to prompt input for retry on device; [25] Not a typewriter
> path : /home/userx
> dev/fset : utilisateurs#utilisateurs
> type : advfs
> advfs id : 0x333106fd.000ed910.1
>
> This is NOT the "out of space" or the "forget to change tape problem .
We
> think the device or the tape could be involved.
>
> My questions are :
> Does TLZ09 correctly work with vdump and 120 m tapes ?
> What's the way to verify the integrity of one tape when some problem
> occurs ?
> what is the life period of a 120 m tape ?
Merry Christmas .....
Chris
-------------------------------------------------------------------------------------
Christophe RETOURNA retourna_at_univ-troyes.fr
Ingenieur de Recherche (33) 03 25 71 76 28
Centre de Ressources Informatiques
Universite de Technologie de Troyes
12 rue Marie Curie
BP 2060
10010 TROYES
Received on Wed Dec 23 1998 - 07:41:42 NZDT