[Search for users] [Overall Top Noters] [List of all Conferences] [Download this site]

Conference cookie::archive_backup

Title:Archive/Backup
Moderator:COOKIE::MHUAIG
Created:Wed Sep 08 1993
Last Modified:Fri Jun 06 1997
Last Successful Update:Fri Jun 06 1997
Number of topics:479
Total number of notes:2283

432.0. "catalog handling functionality ABS >2.1" by ATZIS1::PIEBER (chaos has many faces) Thu Apr 17 1997 11:12

    Hi,
    
    	here is a list of features required for simplified catalog
    handling:
    
    
    1) There should be a way to allow catalog deletion.
    
    	I know, that this may cause inconsistencies with storage classes
    pointing to it, but in case one simply has a typo when creating a
    catalog, currently there is no official way to get rid of it.
    
      --> there is an undocumented(/unsupported?) way to achieve this.
    
    2) The staging attribute should be possible to be toggled even after 
       catalog creation.
     
    	The staging flag is only used during a save operation. I tells,
    whether ABS writes to the indexed files of the catalog directly, or
    first to a sequential file (<Catlg>.STG) and to the indexed files
    (BAOE, BAOE_INSNC) after the actual save operation is complete (staging). 
    
    This means, once the staging update is done, there is no difference for 
    subsequent restores, nor for the next save operation. So it can savely
    be modified. 
    
      --> there is an undocumented(/unsupported?) way to achieve this.
    
    3) there should be a way to pre-expire catalog entries.
    
    	Currently, there is no way in ABS, to eliminate catalog entries,
    unless they reach their expiration date and get removed by the catalog
    cleaner. The other direction, trying to keep entries longer, than
    initially determined, is less important.
    
    
    4) There should be a tool to verify catalog consistency.
    
    	Since a node can crash in the middle of a cleanup operation, or a
    staging update, or the process simply gets deleted, this might leave
    inconsistencies in the catalog. There is no way so far, to tell,
    whether such a case exists, nor can it be corrected. Long retention
    periods will aggravate the situation.
    
    
    Ewald.
T.RTitleUserPersonal
Name
DateLines
432.1Added to requirementsCOOKIE::HELLWIGWed Apr 23 1997 18:3357
           <<< Note 432.0 by ATZIS1::PIEBER "chaos has many faces" >>>
                  -< catalog handling functionality ABS >2.1 >-

    Hi,
    
    	here is a list of features required for simplified catalog
    handling:
    
    
    1) There should be a way to allow catalog deletion.
    
    	I know, that this may cause inconsistencies with storage classes
    pointing to it, but in case one simply has a typo when creating a
    catalog, currently there is no official way to get rid of it.
    
      --> there is an undocumented(/unsupported?) way to achieve this.
    
### We can add this to the requirements list.

    2) The staging attribute should be possible to be toggled even after 
       catalog creation.
     
    	The staging flag is only used during a save operation. I tells,
    whether ABS writes to the indexed files of the catalog directly, or
    first to a sequential file (<Catlg>.STG) and to the indexed files
    (BAOE, BAOE_INSNC) after the actual save operation is complete (staging). 
    
    This means, once the staging update is done, there is no difference for 
    subsequent restores, nor for the next save operation. So it can savely
    be modified. 
    
      --> there is an undocumented(/unsupported?) way to achieve this.

### We can add this to the requirements list.
    
    3) there should be a way to pre-expire catalog entries.
    
    	Currently, there is no way in ABS, to eliminate catalog entries,
    unless they reach their expiration date and get removed by the catalog
    cleaner. The other direction, trying to keep entries longer, than
    initially determined, is less important.

### We can add this to the requirements list... A question for you -
### What would be the criteria to base the selection on?  
    
    4) There should be a tool to verify catalog consistency.
    
    	Since a node can crash in the middle of a cleanup operation, or a
    staging update, or the process simply gets deleted, this might leave
    inconsistencies in the catalog. There is no way so far, to tell,
    whether such a case exists, nor can it be corrected. Long retention
    periods will aggravate the situation.
    
### We can add this to the requirements list.  
    
    Ewald.

432.2simple commands will do...ATZIS2::PIEBERchaos has many facesThu Apr 24 1997 11:4844
    Hi Kim,
    
    > ### What would be the criteria to base the selection on?
    
    For my point of view, it could be as simple as a wildcard specification
    that selects all files desired plus the new specification for an
    expiration date. In that way, one can select what should be covered by
    the command and the new date (absolute or delta) will release the
    objects hit to the catalog cleaner...
    Selction qualifiers /SINCE, /BEFORE are required.
    
    Other ways would come to mind, but would complicate matters:
    
    * as the API now first evaluates a query in ABS LOOKUP first and then
    	starts writing the output, one could keep the selection around
    	until the next command inside ABS and allow the command
    	EXPIRE/Date=<date>. Example session, including error:
    
    	$ ABS
    	ABS> lookup/catalog=pies_Catalog $5$dka300:[Ewald_Pieber...]*.* -
    		/SINCE=<since>/BEFORE=<before>/OBJECT=VMS_FILES
        	.
    		.    maybe tons of output, or /OUT=NL:
    		.
    	ABS> EXPIRE/DATE=1-JAN-1997	! i.e. NOW
    	ABS> exit
    	$ ABS
    	ABS> EXPIRE/DATE=1-JAN-1997	! no cursor, will not work
    	%ABS-F-NOSELECTION, no active selection, use LOOKUP first.
    	ABS> exit
    	$ ............................
    
    * selection by SR: One can use REPORT SAVE_LOG to get the UID for a
    	specific SR and use this in the pre-expiration command.
    * also allow to specify a date/time in the future, thus preventing
    	entries from being cleaned away.
    	==> This might cause headache with regard to MDMS volume
    		retention, since selected files might be on a whole list of
    		volumes and could bring troubles to the previewed pool
    		management.
    
    I think, that's about it.
    
    Ewald.