Q: Data Protector Object Copy with two copies - HP UX

This is a discussion on Q: Data Protector Object Copy with two copies - HP UX ; Hi, We have a configuration for Data Protector that does not work. I think I have made no mistake, so maybe the software cannot do it: We have the desire that every backup object should have a copy. So I ...

+ Reply to Thread
Results 1 to 6 of 6

Thread: Q: Data Protector Object Copy with two copies

  1. Q: Data Protector Object Copy with two copies

    Hi,

    We have a configuration for Data Protector that does not work. I think I have
    made no mistake, so maybe the software cannot do it:

    We have the desire that every backup object should have a copy.

    So I have configured two file juke boxes and media pools, slots, etc.

    Some clients backup to jukebox1, others backup to jukebox2. As soon as a
    backup finished, a post-backup copy (Data Protector 6.0) copies the objects to
    the other jukebox. That works very fine.

    To avoid an overflow of the file jukeboxes, I have a scheduled job that copies
    the objects still protected from the file jukeboxes to LTO3 in a larger
    library, recycling the data protection on the original file object. The
    objects in one file jukebox are strictly copied to a media pool different from
    those for the objects in the other file jukebox.

    The job that copies those "originals" also works fine, but the job that should
    copy the copies in the file library to yet another tape never finds any
    objects to copy. Now the file jukebox for the copies has overflowed as no
    protected objects were moved out of it.

    I've checked the selection criteria several times, but I find
    no mistake there.

    Has anybody successfully implemented a similar system?

    Regards,
    Ulrich

  2. Re: Data Protector Object Copy with two copies

    "Ulrich Windl" wrote in message
    news:87r68u6260.fsf@pc9454.klinik.uni-regensburg.de...
    > Hi,
    >
    > We have a configuration for Data Protector that does not work. I think I
    > have
    > made no mistake, so maybe the software cannot do it:
    >
    > We have the desire that every backup object should have a copy.
    >
    > So I have configured two file juke boxes and media pools, slots, etc.
    >
    > Some clients backup to jukebox1, others backup to jukebox2. As soon as a
    > backup finished, a post-backup copy (Data Protector 6.0) copies the
    > objects to
    > the other jukebox. That works very fine.
    >
    > To avoid an overflow of the file jukeboxes, I have a scheduled job that
    > copies
    > the objects still protected from the file jukeboxes to LTO3 in a larger
    > library, recycling the data protection on the original file object. The
    > objects in one file jukebox are strictly copied to a media pool different
    > from
    > those for the objects in the other file jukebox.
    >
    > The job that copies those "originals" also works fine, but the job that
    > should
    > copy the copies in the file library to yet another tape never finds any
    > objects to copy. Now the file jukebox for the copies has overflowed as no
    > protected objects were moved out of it.
    >
    > I've checked the selection criteria several times, but I find
    > no mistake there.
    >
    > Has anybody successfully implemented a similar system?
    >
    > Regards,
    > Ulrich


    Not implemented anything at all similar, but just some random thoughts,
    which might help, or might lead you up the garden path.

    The only thing similar I have done is to have an "incremental forever"
    (synthetic) backup copy the resultant "synthetic full" object to tape. To
    get this to work I need to have a post-backup consolidation specification
    (triggered from the incremental backup), and then a post-backup copy
    specification triggered from the post-backup consolidation specification.

    When you copy an object, the copy is no longer associated (directly) with
    the original backup session, but instead is associated with the copy
    session. In order copy the copy, you would need to have it select based on
    the automated post backup copy specification, rather than on the original
    backup specification. Unfortunately you don't seem to be able to select an
    copy specification as the source for an automated copy. (Only backup and
    automated consolidation are in the list.).

    The only way I can think of to get around this (apart from directly editing
    the files to see if you can force it in) is to have the copy to the second
    jukebox use a short protection period, so that it expires, and to make the
    two tape copies from the first jukebox.



  3. Re: Data Protector Object Copy with two copies

    "Kilgaard" writes:

    > "Ulrich Windl" wrote in message
    > news:87r68u6260.fsf@pc9454.klinik.uni-regensburg.de...
    >> Hi,
    >>
    >> We have a configuration for Data Protector that does not work. I think I
    >> have
    >> made no mistake, so maybe the software cannot do it:
    >>
    >> We have the desire that every backup object should have a copy.
    >>
    >> So I have configured two file juke boxes and media pools, slots, etc.
    >>
    >> Some clients backup to jukebox1, others backup to jukebox2. As soon as a
    >> backup finished, a post-backup copy (Data Protector 6.0) copies the
    >> objects to
    >> the other jukebox. That works very fine.
    >>
    >> To avoid an overflow of the file jukeboxes, I have a scheduled job that
    >> copies
    >> the objects still protected from the file jukeboxes to LTO3 in a larger
    >> library, recycling the data protection on the original file object. The
    >> objects in one file jukebox are strictly copied to a media pool different
    >> from
    >> those for the objects in the other file jukebox.
    >>
    >> The job that copies those "originals" also works fine, but the job that
    >> should
    >> copy the copies in the file library to yet another tape never finds any
    >> objects to copy. Now the file jukebox for the copies has overflowed as no
    >> protected objects were moved out of it.
    >>
    >> I've checked the selection criteria several times, but I find
    >> no mistake there.
    >>
    >> Has anybody successfully implemented a similar system?
    >>
    >> Regards,
    >> Ulrich

    >
    > Not implemented anything at all similar, but just some random thoughts,
    > which might help, or might lead you up the garden path.
    >
    > The only thing similar I have done is to have an "incremental forever"
    > (synthetic) backup copy the resultant "synthetic full" object to tape. To
    > get this to work I need to have a post-backup consolidation specification
    > (triggered from the incremental backup), and then a post-backup copy
    > specification triggered from the post-backup consolidation specification.
    >
    > When you copy an object, the copy is no longer associated (directly) with
    > the original backup session, but instead is associated with the copy
    > session. In order copy the copy, you would need to have it select based on
    > the automated post backup copy specification, rather than on the original
    > backup specification. Unfortunately you don't seem to be able to select an
    > copy specification as the source for an automated copy. (Only backup and
    > automated consolidation are in the list.).
    >
    > The only way I can think of to get around this (apart from directly editing
    > the files to see if you can force it in) is to have the copy to the second
    > jukebox use a short protection period, so that it expires, and to make the
    > two tape copies from the first jukebox.


    Hi!

    You just seem to have explained what's going on. Unfortunately the first copy
    to tape also recycles the objects in the first jukebox after copy, so a second
    copy from the first jukebox won't work (as I'm only copying protected
    objects). Despite of that we also copy the object protection, so that
    assigning a short protection is not the way to go. Unfortunately.

    As you described, the bug seems to be that the copy is associated with the
    copy specification and no with the original backup specification. So my select
    for objects (2nd copy, copy of copy) goes into nirvana.

    If I have more nerves than now, I might file a support request for HP, but I
    can't stand it at the moment (Usually once it turns out that the problem is
    really a problem, HP support becomes very silent while you are waiting for a
    solution).

    Regards,
    Ulrich

  4. Re: Data Protector Object Copy with two copies

    "Ulrich Windl" wrote in message
    news:87r68s106l.fsf@pc9454.klinik.uni-regensburg.de...
    > "Kilgaard" writes:
    >
    >> "Ulrich Windl" wrote in message
    >> news:87r68u6260.fsf@pc9454.klinik.uni-regensburg.de...
    >>> Hi,
    >>>
    >>> We have a configuration for Data Protector that does not work. I think I
    >>> have
    >>> made no mistake, so maybe the software cannot do it:
    >>>
    >>> We have the desire that every backup object should have a copy.
    >>>
    >>> So I have configured two file juke boxes and media pools, slots, etc.
    >>>
    >>> Some clients backup to jukebox1, others backup to jukebox2. As soon as a
    >>> backup finished, a post-backup copy (Data Protector 6.0) copies the
    >>> objects to
    >>> the other jukebox. That works very fine.
    >>>
    >>> To avoid an overflow of the file jukeboxes, I have a scheduled job that
    >>> copies
    >>> the objects still protected from the file jukeboxes to LTO3 in a larger
    >>> library, recycling the data protection on the original file object. The
    >>> objects in one file jukebox are strictly copied to a media pool
    >>> different
    >>> from
    >>> those for the objects in the other file jukebox.
    >>>
    >>> The job that copies those "originals" also works fine, but the job that
    >>> should
    >>> copy the copies in the file library to yet another tape never finds any
    >>> objects to copy. Now the file jukebox for the copies has overflowed as
    >>> no
    >>> protected objects were moved out of it.
    >>>
    >>> I've checked the selection criteria several times, but I find
    >>> no mistake there.
    >>>
    >>> Has anybody successfully implemented a similar system?
    >>>
    >>> Regards,
    >>> Ulrich

    >>
    >> Not implemented anything at all similar, but just some random thoughts,
    >> which might help, or might lead you up the garden path.
    >>
    >> The only thing similar I have done is to have an "incremental forever"
    >> (synthetic) backup copy the resultant "synthetic full" object to tape. To
    >> get this to work I need to have a post-backup consolidation specification
    >> (triggered from the incremental backup), and then a post-backup copy
    >> specification triggered from the post-backup consolidation specification.
    >>
    >> When you copy an object, the copy is no longer associated (directly) with
    >> the original backup session, but instead is associated with the copy
    >> session. In order copy the copy, you would need to have it select based
    >> on
    >> the automated post backup copy specification, rather than on the original
    >> backup specification. Unfortunately you don't seem to be able to select
    >> an
    >> copy specification as the source for an automated copy. (Only backup and
    >> automated consolidation are in the list.).
    >>
    >> The only way I can think of to get around this (apart from directly
    >> editing
    >> the files to see if you can force it in) is to have the copy to the
    >> second
    >> jukebox use a short protection period, so that it expires, and to make
    >> the
    >> two tape copies from the first jukebox.

    >
    > Hi!
    >
    > You just seem to have explained what's going on. Unfortunately the first
    > copy
    > to tape also recycles the objects in the first jukebox after copy, so a
    > second
    > copy from the first jukebox won't work (as I'm only copying protected
    > objects). Despite of that we also copy the object protection, so that
    > assigning a short protection is not the way to go. Unfortunately.
    >
    > As you described, the bug seems to be that the copy is associated with the
    > copy specification and no with the original backup specification. So my
    > select
    > for objects (2nd copy, copy of copy) goes into nirvana.
    >
    > If I have more nerves than now, I might file a support request for HP, but
    > I
    > can't stand it at the moment (Usually once it turns out that the problem
    > is
    > really a problem, HP support becomes very silent while you are waiting for
    > a
    > solution).
    >
    > Regards,
    > Ulrich


    I don't see the problem is with using a shorter protection on the second
    jukebox *if* you can make both tape copies from the first jukebox.

    To make two copies from the same jukebox you just need to make the first
    copy *without* the recycle protections flag set, and then make the second
    one *with* it set. You will need to be careful with the scheduling and
    filtering to make sure that you get the same set of sessions in each copy.

    If you are copying out once per day then you should be able to do something
    like:

    3am: copy sessions from 48 hours ago with duration 24 hours no recycle
    4am: copy sessions from 49 hours ago with duration 24 hours recycle

    Your post-backup copy just sets the protection to 49 (or 73) hours and it
    will expire around the time the second copy is made (or a day later if you
    want to be safer, which I think you do).



  5. Re: Data Protector Object Copy with two copies

    "Kilgaard" writes:

    > "Ulrich Windl" wrote in message
    > news:87r68s106l.fsf@pc9454.klinik.uni-regensburg.de...
    >> "Kilgaard" writes:
    >>
    >>> "Ulrich Windl" wrote in message
    >>> news:87r68u6260.fsf@pc9454.klinik.uni-regensburg.de...
    >>>> Hi,
    >>>>
    >>>> We have a configuration for Data Protector that does not work. I think I
    >>>> have
    >>>> made no mistake, so maybe the software cannot do it:
    >>>>
    >>>> We have the desire that every backup object should have a copy.
    >>>>
    >>>> So I have configured two file juke boxes and media pools, slots, etc.
    >>>>
    >>>> Some clients backup to jukebox1, others backup to jukebox2. As soon as a
    >>>> backup finished, a post-backup copy (Data Protector 6.0) copies the
    >>>> objects to
    >>>> the other jukebox. That works very fine.
    >>>>
    >>>> To avoid an overflow of the file jukeboxes, I have a scheduled job that
    >>>> copies
    >>>> the objects still protected from the file jukeboxes to LTO3 in a larger
    >>>> library, recycling the data protection on the original file object. The
    >>>> objects in one file jukebox are strictly copied to a media pool
    >>>> different
    >>>> from
    >>>> those for the objects in the other file jukebox.
    >>>>
    >>>> The job that copies those "originals" also works fine, but the job that
    >>>> should
    >>>> copy the copies in the file library to yet another tape never finds any
    >>>> objects to copy. Now the file jukebox for the copies has overflowed as
    >>>> no
    >>>> protected objects were moved out of it.
    >>>>
    >>>> I've checked the selection criteria several times, but I find
    >>>> no mistake there.
    >>>>
    >>>> Has anybody successfully implemented a similar system?
    >>>>
    >>>> Regards,
    >>>> Ulrich

    [...]
    > I don't see the problem is with using a shorter protection on the second
    > jukebox *if* you can make both tape copies from the first jukebox.


    The idea is about automatism: Assuming the first jukebox fails after an object
    had been stored there (and after it was copied to jukebox #2), there would not
    be any objects on tape then.

    >
    > To make two copies from the same jukebox you just need to make the first
    > copy *without* the recycle protections flag set, and then make the second
    > one *with* it set. You will need to be careful with the scheduling and
    > filtering to make sure that you get the same set of sessions in each copy.


    So if the first copy to tape fails, the second will recycle the objects,
    preventing two copies of every object. (If my pattern worked, the first copy
    would still be retried nect time unless the object's protection had expired
    meanwhile)

    >
    > If you are copying out once per day then you should be able to do something
    > like:
    >
    > 3am: copy sessions from 48 hours ago with duration 24 hours no recycle
    > 4am: copy sessions from 49 hours ago with duration 24 hours recycle


    Things are a bit more complicated in reality as we work with several different
    media Pools on the file jukeboxes and the LTO3 tapes. That is, each media pool
    on the file jukebox corresponds to a media pool on the tape library.

    >
    > Your post-backup copy just sets the protection to 49 (or 73) hours and it
    > will expire around the time the second copy is made (or a day later if you
    > want to be safer, which I think you do).


    Even that is more complicated: For some systems we use 9 backup levels in a
    somewhat nested way, and every level uses a data protection that depends on
    the level and the time and level of the next scheduled backup for the same
    object. (In simple words: If you do a level-8-backup four times a week, and
    you do an incremental/differential daily, there little need to keep the daily
    incremental much longer than two days. Or if you do a level-1 on Monday, and a
    full on tuesday, you can discard the level-1 rather soon)

    Regards,
    Ulrich

  6. Re: Data Protector Object Copy with two copies

    "Ulrich Windl" wrote in message
    news:87hc9iy8ta.fsf@pc9454.klinik.uni-regensburg.de...
    > "Kilgaard" writes:
    >
    >> "Ulrich Windl" wrote in message
    >> news:87r68s106l.fsf@pc9454.klinik.uni-regensburg.de...
    >>> "Kilgaard" writes:
    >>>
    >>>> "Ulrich Windl" wrote in message
    >>>> news:87r68u6260.fsf@pc9454.klinik.uni-regensburg.de...
    >>>>> Hi,
    >>>>>
    >>>>> We have a configuration for Data Protector that does not work. I think
    >>>>> I
    >>>>> have
    >>>>> made no mistake, so maybe the software cannot do it:
    >>>>>
    >>>>> We have the desire that every backup object should have a copy.
    >>>>>
    >>>>> So I have configured two file juke boxes and media pools, slots, etc.
    >>>>>
    >>>>> Some clients backup to jukebox1, others backup to jukebox2. As soon as
    >>>>> a
    >>>>> backup finished, a post-backup copy (Data Protector 6.0) copies the
    >>>>> objects to
    >>>>> the other jukebox. That works very fine.
    >>>>>
    >>>>> To avoid an overflow of the file jukeboxes, I have a scheduled job
    >>>>> that
    >>>>> copies
    >>>>> the objects still protected from the file jukeboxes to LTO3 in a
    >>>>> larger
    >>>>> library, recycling the data protection on the original file object.
    >>>>> The
    >>>>> objects in one file jukebox are strictly copied to a media pool
    >>>>> different
    >>>>> from
    >>>>> those for the objects in the other file jukebox.
    >>>>>
    >>>>> The job that copies those "originals" also works fine, but the job
    >>>>> that
    >>>>> should
    >>>>> copy the copies in the file library to yet another tape never finds
    >>>>> any
    >>>>> objects to copy. Now the file jukebox for the copies has overflowed as
    >>>>> no
    >>>>> protected objects were moved out of it.
    >>>>>
    >>>>> I've checked the selection criteria several times, but I find
    >>>>> no mistake there.
    >>>>>
    >>>>> Has anybody successfully implemented a similar system?
    >>>>>
    >>>>> Regards,
    >>>>> Ulrich

    > [...]
    >> I don't see the problem is with using a shorter protection on the second
    >> jukebox *if* you can make both tape copies from the first jukebox.

    >
    > The idea is about automatism: Assuming the first jukebox fails after an
    > object
    > had been stored there (and after it was copied to jukebox #2), there would
    > not
    > be any objects on tape then.
    >


    True, but I guess I would consider a jukebox failure to be something
    requiring special attention and would then manually perform the copy to
    tape.

    >>
    >> To make two copies from the same jukebox you just need to make the first
    >> copy *without* the recycle protections flag set, and then make the second
    >> one *with* it set. You will need to be careful with the scheduling and
    >> filtering to make sure that you get the same set of sessions in each
    >> copy.

    >
    > So if the first copy to tape fails, the second will recycle the objects,
    > preventing two copies of every object. (If my pattern worked, the first
    > copy
    > would still be retried nect time unless the object's protection had
    > expired
    > meanwhile)
    >


    Once again, I would consider a failure to copy to tape (of any sessions) to
    be something that requires special attention and possible manual recovery.
    If the first attempt to copy to tape fails, you must be more optimistic than
    me to expect the second attempt of it to work without needing manual
    intervention.

    >>
    >> If you are copying out once per day then you should be able to do
    >> something
    >> like:
    >>
    >> 3am: copy sessions from 48 hours ago with duration 24 hours no recycle
    >> 4am: copy sessions from 49 hours ago with duration 24 hours recycle

    >
    > Things are a bit more complicated in reality as we work with several
    > different
    > media Pools on the file jukeboxes and the LTO3 tapes. That is, each media
    > pool
    > on the file jukebox corresponds to a media pool on the tape library.


    Adds more specifications (one for each media pool) and might require some
    careful scheduling (if 15 minutes is not fine grained enough) but otherwise
    does not significantly alter the problem.

    >
    >>
    >> Your post-backup copy just sets the protection to 49 (or 73) hours and it
    >> will expire around the time the second copy is made (or a day later if
    >> you
    >> want to be safer, which I think you do).

    >
    > Even that is more complicated: For some systems we use 9 backup levels in
    > a
    > somewhat nested way, and every level uses a data protection that depends
    > on
    > the level and the time and level of the next scheduled backup for the same
    > object. (In simple words: If you do a level-8-backup four times a week,
    > and
    > you do an incremental/differential daily, there little need to keep the
    > daily
    > incremental much longer than two days. Or if you do a level-1 on Monday,
    > and a
    > full on tuesday, you can discard the level-1 rather soon)
    >


    It's amazing the number of backup plans I have seen that don't understand
    that. They quite happily make a full on Sunday and incremental throughout
    the week, only to throw away the full on about Wednesday, and wonder why
    they can't do a recovery on Friday.

    But, yes, given this system, setting a reasonable protection on the copy to
    jukebox 2 would be challenging.



+ Reply to Thread