ADSM-L

Re: Archives and management classes for directories

1999-05-20 12:19:06
Subject: Re: Archives and management classes for directories
From: Paul Zarnowski <vkm AT CORNELLC.CIT.CORNELL DOT EDU>
Date: Thu, 20 May 1999 12:19:06 -0400
At 10:09 AM 5/20/99 +0200, Reinhard Mersch wrote:
>Trevor,
>
>let me explain my wholehearted assistance to your complaints. In March,
>Paul Zarnowski already complained about it (Subject: Archive retention for
>directories). Paul, has anything happend in between?

Yes, just this week the kind folks in the Support Center have informed me
that the set of related Archive problems will be fixed.  There are
primarily two APARs related to this.  The first, IX89638 has already been
appended to this list in an earlier response from Rejean Larivee (thank
you, Rejean).  The second is IX85975, which I will append at the end of
this e-mail.  With these fixes, I believe that ADSM Archive will behave
correctly, allowing the type of archival usage that I and others have
discussed here.  I was told that the fixes will be included in the next fix
level.

>In my case, this new "feature" of ADSM V.3 prevents at least one of my
>clients from being upgraded to version 3. This client archives about 200
>files per day, using seperate archive commands. The files are archived for
>30 days. With version 3, the 5 directories above these files would also be
>archived, and they would be archived for 10 years. I would end up with
>3660000 archived directories (5 directories for each of the 200 files per
>day archived for 3660 days), which largely are identical (the archived
>files all come from the same directory), and which are never needed.
>
>This way, my perpetual struggle to keep the ADSM data base at a manageable
>size is condemned to fail.

Exactly right - this problem would make itself evident over time.  But -
this is not a "feature".  IBM now recognizes it as a bug, and will be
providing fixes.  There are fixes to both the server and the client.

As a side note, I would encourage folks to call problems into the Support
Center when they encounter them.  I know this can take precious time, but
it really does help to improve the product not just for you, but for the
entire community of ADSM users.  When we first called in this problem, back
in December, I was told that there were APARs for this that were closed as
SUG (Suggestion).  I suspect that IBM did not realize the full impact of
this bug at first, and might have realized it sooner if more folks had
called it in.  This is just speculation on my part, of course.  I know it
can be frustrating trying to convince someone else that there is a bug when
it clear to you, having just experienced it.  However, I have found that
(with some patience), once they understand the problem they take action to
fix it.

Possibly some good has come out of this - since we could not wait for the
fix for this problem for one of our larger users, we decided to explore
converting them to use HSM.

..Paul
---
Item IX85975
Item IX85975

  APAR Identifier ...... IX85975      Last Changed ........ 99/04/26
  DELETE ARCHIVE DELETES THE FILES AND DIRECTORIES BELOW THE
  SPECIFIED DIRECTORY, BUT NOT THE SPECIFIED DIRECTORY ITSELF.

  Symptom ...... IN INCORROUT         Status ........... CLOSED  PER
  Severity ................... 3      Date Closed ......... 99/01/18
  Component .......... 565511904      Duplicate of ........
  Reported Release ......... 31A      Fixed Release ............ 999
  Component Name ADSM AIX V4 CLI      Special Notice
  Current Target Date ..99/05/08      Flags
  SCP ................... AIXRSC
  Platform ............ AIX

  Status Detail: Not Available

  PE PTF List:

  PTF List:
  Release 31A   : PTF not available yet
  Release 31B   : PTF not available yet


  Parent APAR:    IX85967
  Child APAR list: IC23738 PQ26379


  ERROR DESCRIPTION:
  DELETE ARCHIVE deletes archive versions of files and directories
  below the specified directory, but does not delete the specified
  directory itself. The only way to delete the specified directory
  is to delete the next highest-level directory, but this would
  also cause peer files and directories of the specified directory
  to also be deleted. For example, suppose the following directory
  structure is archived:

  /TopDir/Dir1/File1
  /TopDir/Dir1/File2
  /TopDir/Dir2/File3
  /TopDir/File4

  The command "DELETE ARCHIVE /TopDir/Dir1/ -SUBDIR=YES" would
  delete the archived versions of File1 and File2, but Dir1 itself
  would not be deleted.

  The only way to delete Dir1 is to issue the command
  "DELETE ARCHIVE /TopDir/ -SUBDIR=YES", except this has the
  undesired side-effect of deleting *all* the files and
  directories under TopDir, including Dir2 and its files, and
  File4.

  The DELETE ARCHIVE command should delete not only the files
  and subdirectories in the specified directory, but the specified
  directory itself.


  LOCAL FIX:
  Currently the user can delete the directory using -PICK with the
  DELETE ARCHIVE command, but that capability is a defect (APAR
  IC22894) that will be removed from the product in the next PTF.
  Note that using the -PICK option won't work in scripts.


  PROBLEM SUMMARY:
  ****************************************************************
  *USERS AFFECTED: All Platforms                                 *
  ****************************************************************
  *PROBLEM DESCRIPTION: Delete Archive -subdir=yes, the files and*
  *                     directories below the specified directory*
  *                     are deleted, but not the directory itself*
  ****************************************************************


  PROBLEM CONCLUSION:
  When archive_delete is performed, dsmc queries the server with
  the required filespec and then builds list of files/directories
  to be deleted. The server returns ONLY "childs" of the specified
  directory, not directory itself. It must be added with the help
  of one more query.
  Now, if DELETE ARCHIVE -SUBDIR=YES is issued for a directory,
  that directory as well as its files and subdirectories will be
  deleted.


  TEMPORARY FIX:


  COMMENTS:


  MODULES/MACROS:   EXE      DSMC
                    EXE      DSM


  SRLS:      NONE


  RTN CODES:


  CIRCUMVENTION:


  MESSAGE TO SUBMITTER:
  2/25/99, ml, added to conclusion
---
=========================================================================
<Prev in Thread] Current Thread [Next in Thread>