Re: [Bacula-users] Cleaning Jobs

[Arno L. <al...@it...>]

Hi,

On 2/22/2007 8:39 AM, Pierre Bernhardt wrote:
> Arno Lehmann schrieb:
...
>>Autochangers can usually manage cleaning on their own - you dedicate one 
>>slot to the cleaning tape, and everything else happens automatically. I 
>>don't know if tis interferes witch Baculas operation, though - I don't 
>>operate an autochanger like that.
> 
> Usually only, if this is a high expensive tape library. I have bought an
> Libra-8 8 cartrige DDS-3 and an 6 cartrige DDS-2 HP libraries.
> Both libs have no barcode reade.
> It is possible that both can use a cleaning cartrige.
> 
> But update slots scan command after cartrige change of the other ones
> are not a good Idea because the cleaning cartrige is loaded, too.
> At this time the tape is loaded but will eject by the drive but not unload.
> And bacula don't eat the cartrige and is still waiting for the
> data on it (but could never read them)

Right. I run a 6-slot HP thingie, too, and given the limited slots I 
don't keep a cleaning tape in there. It's easier to have a spare 
magazine with a cleaning tape and, when drive cleaning becomes 
necessary, to insert it and manually trigger a cleaning cycle. Given the 
low frequency needed for drive cleaning, that makes less magazine 
changes than running with five tapes only.

> 
>>>Is this supported in Bacula, or is it just easier/better to have a quick
>>>and dirty script run from cron, that uses mtx to load and then unload
>>>the tape???
> 
> If I use mtx bacula will not see what I did with the other cartriges.
> So I got problems because bacula has other information stored.

I suggest to not clean the tape from Bacula. If you *really* want a 
cleaning job, make that an admin job which calls a run before script to 
unmount and unload the drive, load the cleaning tape, note the cleaning 
tape usage in a database, and load and mount the previous tape again.

I never saw the need for that, though, so I don't have a script at hand :-)


>>There is a feature request open regarding better tape cleaning 
>>management... until that is implemented, I do that manually.
> 
> At first I need a option so bacula will never use a proibited slot.

Modify your mtx-changer accordingly.

> At the moment I know only, that I can use Cleaning prefix and than
> add the Cartrige manually by add command. But I have not test this
> because the update slot scan feature will check the slot again.

The cleaning prefix is only useful with barcode readers.

> Only modifying the mtxchanger script is a workaround, but then
> I must modify it on a very dirty way.

That doesn't have to be dirty IMO.

>>Whenever there is a tapealert (which you can get into Baculas job 
>>report) or a drive failure, it's time for manual intervention. Not 
>>really a serious problem, because in a clean environment you won't have 
>>to clean your drives very often. And you wouldn't want to operate a 
>>modern tape drive in a dusty environment :-)
> 
> Ok. Some points of a feature for next bacula:
> 
> Here my feature request for a cleaning enhacement:

If you really want that to be a feature request you should rewrite it 
into the format suggested by Kern. You can find that on the project web 
site, I believe.

> 1. It should be possible to declare for one or more cleaning slots
>    in a library.

That would be done in mtx-changer, possibly configurable. This would be 
possible rather easily, I think.

> 2. A max. cleaning procedures for a cleaning tape should declared.
> 3. The cleaning procedure should only arrange by bacula (count the
>    cleaning procedures)

This is encompassed by the existing feature request, IMO.

> 4. The cleaning procedure should be declared by a job.

No. You really don't want to clean a drive unless the drive itself asks 
you to.

> 5. The job should be run by a request (tape alert, run command,...)
> 6. The cleaning job should be abort if the number of cleaning requests
>    are arrived (message or other)

I don't think that we need extra job types for drive cleaning. I rather 
think that an external script would be best - think of mtx-changer for 
tape movement - or, as a first step, a simple operator intervention request.

> 7. update slots scan should never scan the defined cleaning slots.

Item 1.

> 8. cleaning slots should never requests for reading/writing.

Item 1.

> What did the tape if I do cleaning manually:
> (Cleaning cartrige in slot 8)
> 1. mtx load 8
> The changer loads the cartrige in the drive
> The drive inject the cartrige and cleaning process is startet.
> The drive ejects the cartrige and know its different:
> Libra-8: mtx unload 8 is needed because the drive eject don't
>    unload the cartrige back in slot.
> HP 12000: The drive eject the cartrige and the changer unloads
>    the cartrige automatically back in the slot (6)
>    (I'm not reallly shure. didn't use the library any more)
> 

Should work like that. Mine does, at least :-)

> Discussion is welcome :-)

First, drive cleaning is not a daily operation, so the actual loading of 
the cleaning tape could be done manually, as a first step.

The more important step towards integrated drive cleaning is to 
interface that into the normal tape operation. We will need job pausing 
for this. Just imagine that you have a really huge job running, and 
after the first few blocks written to tape the drive wants to be 
cleaned. Today, you'd have to wait until that tape is no longer busy, 
i.e. until all jobs to that tape are finished, or until a tape change is 
required. Being able to pause the job, unmount the tape, clean the 
drive, mount the tape again and continue the job would be a major 
improvement.

Arno

> MfG...

Also mfG ;-)

> Pierre Bernhardt
> 
> 
> -------------------------------------------------------------------------
> Take Surveys. Earn Cash. Influence the Future of IT
> Join SourceForge.net's Techsay panel and you'll get the chance to share your
> opinions on IT & business topics through brief surveys-and earn cash
> http://www.techsay.com/default.php?page=join.php&p=sourceforge&CID=DEVDEV
> _______________________________________________
> Bacula-users mailing list
> Bac...@li...
> https://lists.sourceforge.net/lists/listinfo/bacula-users

-- 
IT-Service Lehmann                    al...@it...
Arno Lehmann                  http://www.its-lehmann.de