On Tuesday 24 November 2009 22:15:35 Martin Simmons wrote:
> >>>>> On Sun, 22 Nov 2009 12:16:33 +0100, Kern Sibbald said:
> >
> > On Friday 20 November 2009 14:44:54 Martin Simmons wrote:
> > > It there a technical reason why the NextPool directive is part of the
> > > pool at all?
> >
> > It was not really a technical reason, but David Boyes suggested doing it
> > that way, possibly because that is how TSM does it.  At the time, I
> > wasn't really sure if that was the best way, but today, I am sure that it
> > was the best way to do what we want (good suggestion David) -- see below
> > for why ...
> >
> > > When I first heard about migration and copy jobs, I expected to
> > > find something like a TargetPool directive in the job definition, so
> > > that it can be varied for each job.
> >
> > Although it would slightly facilitate the configuration in a few cases by
> > allowing a target pool in the Job, it would have some negative
> > consequences, namely that if like many users you separate Full, Diff, Inc
> > into separate pools, it would vastly complicate configurations --
> > requiring schedule overrides and such.
>
> OK, that makes sense.
>
> > However, Pool and Storage schedule overrides have turned out to be a real
> > pain to get right for a number of reasons -- the most obvious is if a job
> > is automatically upgraded, in general, the schedule overrides will get
> > botched up.  If I can, I intend to remove Pool and Storage overrides in
> > the schedules.
> >
> > I think the best place for the Storage definition (suggested by David) is
> > in the Pool, and likewise the best place for the output Storage is
> > through the Next Pool in the Pool definition.  Unless I am mistaken, it
> > is just as easy to change the Pool definition in the Job as it is to
> > change what would be Target Pool in the Job.
>
> Yes, it is easy to change the Pool in the Job definition, but it seems
> rather hackish to have dummy Pool definitions just so they can contain the
> desired Next Pool.

I completely agree, but I haven't yet see a nice overall solution.

>
> > Anyway, that is how I currently see it.  If I could see a real practical
> > example why a Target Pool is necessary, I would consider it.
>
> Maybe Migrate and Copy have slightly different requirements?  Or perhaps
> the current design is targeted at scheduled migration and copying, whereas
> some people want to use it for ad hoc copying?

Yes, they were both definitely targeted at scheduled job.  For ad hoc stuff, I 
think someone should suggest a run command nextpool override or something 
like that.  Those kinds of additions don't ususally create too many problems 
aside from the problem of making sure command overrides always work ...

>
> Here is an example: suppose you want to make regular backups of all data
> for disaster recovery, but also copy some of the jobs to special archive
> pools. There might be a regulatory requirement to archive emails plus a
> security requirement to archive traffic logs.  The current design makes it
> difficut to use a single pool for the disaster recovery backups.

I am open to suggestions (provided they don't created more problems long 
term).

Regards,

Kern

>
> > One thing that a lot of users may not know is that the Next Pool can
> > point to the current Pool.  In the original Migrate implementation this
> > could cause a deadlock reading and writing on the same Volume, but in
> > currently released versions, regardless of the Pool specification and the
> > Next Pool specification is, Bacula should never attempt to read and write
> > on the same Volume.  If it does, then I would like to hear about it since
> > I consider that a bug (actually it will be some race condition I didn't
> > foresee in the code).
>
> Yes, that's nice.
>
> __Martin
>
> > Kern
> >
> > > Someone already had to hack around this with a dummy pool definition:
> > >
> > > http://thread.gmane.org/gmane.comp.sysutils.backup.bacula.devel/14084
> > >
> > > __Martin
> > >
> > > >>>>> On Fri, 20 Nov 2009 13:45:06 +0100 (CET), Kern Sibbald said:
> > > >
> > > > Hello,
> > > >
> > > > If the only reason is to prevent a deadlock then this override is not
> > > > necessary since Bacula (at some version, I think it was 2.4.0) will
> > > > not use the same volume for reading and writing.
> > > >
> > > > If there is some other good reason, please remove the comments about
> > > > deadlock and resubmit, and I think you can remove any workarounds you
> > > > implemented to prevent deadlocks.
> > > >
> > > > Kern
> > > >
> > > > > Item ?:       Allow Schedule Resource to override NextPool
> > > > >
> > > > > Date: 18 November 2009
> > > > >
> > > > > Origin:       Jim Barber. [email protected]
> > > > >
> > > > > Status:       New request
> > > > >
> > > > > What: Allow the Schedule resource to define a NextPool= statement
> > > > >       to override the NextPool statement of the pool defined in the 
> > > > > job.
> > > > >
> > > > > Why:  I have an incremental pool that each week gets consolidated
> > > > > into a full pool via a VirtualFull job. The NextPool directive of
> > > > > the incremental pool defines the location of the full pool.
> > > > >
> > > > >       The following week, the next VirtualFull backup will run. It 
> > > > > will
> > > > >       read the previous full backups and incremental backups since 
> > > > > then,
> > > > >       to create new full backups. It is important that the VirtualFull
> > > > >       backup does not try to write to the same tape that the previous
> > > > >       weeks full backup wrote to and left in Append status. Otherwise
> > > > > you could end up with the one tape trying to be read and written
> > > > > and dead-lock.
> > > > >
> > > > >       At the moment I have a hack to get around this. An Admin job 
> > > > > calls
> > > > >       an external command that runs a SQL update to find any tapes in
> > > > > the full pool with an APPEND status and change it to USED. This
> > > > > runs after the full backups have been done.
> > > > >
> > > > >       Instead I'd like to create two full pools. One for even weeks 
> > > > > and
> > > > >       one for odd weeks of the year. That way, even week virtual full
> > > > >       backups could consolidate odd week virtual full backups with the
> > > > >       latest incremental backups. And the odd week virtual full 
> > > > > backups
> > > > >       could consolidate the even week full backups with the latest
> > > > >       incremental backups.
> > > > >
> > > > >       The trouble is that the Incremental pool can only define one 
> > > > > Next
> > > > >       Pool. I can't have it toggle the Next Pool directive from odd to
> > > > >       even, week to week. Unless I could override it from the 
> > > > > schedule.
> > > > >
> > > > >       Doing that would mean I could ditch my SQL hack to manipulate 
> > > > > the
> > > > >       tape status. It will also be less wasteful of tapes, since I 
> > > > > won't
> > > > >       have partially filled USED tapes throughout my library.
> > > > >
> > > > >       There are possibly many uses for such an override that I haven't
> > > > >       thought about.
> > > > >
> > > > > -------------------------------------------------------------------
> > > > >---- ------- Let Crystal Reports handle the reporting - Free Crystal
> > > > > Reports 2008 30-Day
> > > > > trial. Simplify your report design, integration and deployment -
> > > > > and focus on
> > > > > what you do best, core application coding. Discover what's new with
> > > > > Crystal Reports now.  http://p.sf.net/sfu/bobj-july
> > > > > _______________________________________________
> > > > > Bacula-devel mailing list
> > > > > [email protected]
> > > > > https://lists.sourceforge.net/lists/listinfo/bacula-devel
> > > >
> > > > Best regards, Kern
> > > >
> > > > ---------------------------------------------------------------------
> > > >---- ----- Let Crystal Reports handle the reporting - Free Crystal
> > > > Reports 2008 30-Day trial. Simplify your report design, integration
> > > > and deployment - and focus on what you do best, core application
> > > > coding. Discover what's new with Crystal Reports now. 
> > > > http://p.sf.net/sfu/bobj-july
> > > > _______________________________________________
> > > > Bacula-devel mailing list
> > > > [email protected]
> > > > https://lists.sourceforge.net/lists/listinfo/bacula-devel
> > >
> > > -----------------------------------------------------------------------
> > >---- --- Let Crystal Reports handle the reporting - Free Crystal Reports
> > > 2008 30-Day trial. Simplify your report design, integration and
> > > deployment - and focus on what you do best, core application coding.
> > > Discover what's new with Crystal Reports now. 
> > > http://p.sf.net/sfu/bobj-july
> > > _______________________________________________
> > > Bacula-devel mailing list
> > > [email protected]
> > > https://lists.sourceforge.net/lists/listinfo/bacula-devel



------------------------------------------------------------------------------
Let Crystal Reports handle the reporting - Free Crystal Reports 2008 30-Day 
trial. Simplify your report design, integration and deployment - and focus on 
what you do best, core application coding. Discover what's new with
Crystal Reports now.  http://p.sf.net/sfu/bobj-july
_______________________________________________
Bacula-devel mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/bacula-devel

Reply via email to