[Gllug] flushing tape drive cache - SOLVED

[Alain Williams]

On Mon, Mar 27, 2006 at 02:45:03PM +0100, Ryan Cartwright wrote:
> Hi all,
> 
> -- egg on face - long explanation follows ---
> 
> further to my earlier post - the drive cache was not the problem - the 
> symlink was.
> 
> * The drive is an ATAPI drive and it is the first IDE/ATA device
> * /dev/tape is symlinked to /dev/hda
> * `ls -lh /dev/hda` revealed this..
> 
> -rw-r--r--    1 root     root          43G 2006-03-25 01:32 /dev/hda

Eeek! That implies that you have a *huge* RFS[**], I must admit that I don't
like that, I always create the RFS as small[%%] partition, the point being
that if something horrible happens a small (& hopefully quiescent RFS[$$]) will
not get corrupted.

[**] OK: that is what most linux installers do be default as it is the no brainer.

[%%] About 1GB these days, which is obscenely big anyway, especially since I have
/usr as another partition.

[$$] which is why /tmp and /home are in other partitions

> note the lack of the c at the front there - or indeed the device 
> numbers. This is a local file and not a device at all....all my backups 
> have been going to a local file and not to a tape (which I have been 
> carrying around in my bag!!!!). We hadn't spotted this because whenever 
> we retrieved a file it was from last night's backup and thus the file 
> /dev/hda was able to deliver the file we wanted to retrieve. We 
> bascially only noticed this because we were retrieving a file someone 
> deleted on Friday and they wanted it back from Thursday's backup. 
> (fortunately they had mailed someone else a copy for comments and got it 
> back from them!).
> 
> One of the other servers running here (which is running correctly) has 
> the ide-tape module loaded which presents the ide tape drive as /dev/ht0 
> and `ls -lh /dev/ht0` gives
> 
> crw-------    1 root     root      37,   0 2003-10-02 21:45 /dev/ht0
> 
> Thus `mt -f /dev/tape status` on *that* machine gives the status of the 
> tape drive (/dev/tape symlinks to /dev/ht0 on there).
> 
> thanks for the suggestions anyway. I currently have people checking the 
> other servers (not many) for the same or similar setup errors.
> 
> Ryan
> BTW - if anyone is interested mt does have a drvbuffer option which 
> works on SCSI drives.
> 
> `mt -f <device> drvbuffer 0` switches off the drive's buffer and 1 will 
> switch it on again.
> 
> Not that it would have helped me in this case as it's an ATAPI drive.
> -- 
> Gllug mailing list  -  Gllug at gllug.org.uk
> http://lists.gllug.org.uk/mailman/listinfo/gllug

-- 
Alain Williams
Parliament Hill Computers Ltd.
Linux Consultant - Mail systems, Web sites, Networking, Programmer, IT Lecturer.
+44 (0) 787 668 0256  http://www.phcomp.co.uk/

#include <std_disclaimer.h>
-- 
Gllug mailing list  -  Gllug at gllug.org.uk
http://lists.gllug.org.uk/mailman/listinfo/gllug