In <bitnet.adsm-l%ADSM-L%95102311033336 AT VM.MARIST DOT EDU>, ramke AT 
vnet.ibm DOT com (Frank Ramke) writes:
>> ...
>> Is there some option/parameter to use to have busy files "auto-skipped"
>> without actually requiring a reply to a prompt?
>
>Jeff, currently no auto-skip for locked files during write.
>
>We are looking at this and other restore problems that skipping/continuation
>of the restore would be appropriate. Currently, we are thinking of putting
>an entry in the error log and a failure noted in the restore statistics
>for each file that is not recovered. Is this enough?
>
>Do you, or others in this list, have additional ideas?
>
>Frank Ramke

Messages are not enough if one is left with lots of manual work to
do to complete the restore.  Please design with BIG servers in mind!

Consider a restore of 100k files which is 99.8% effective.  It leaves
200 files to be fixed up in some way.  I suggest the following -

1). Add an option '-FailuresFile=<filespec>'.  Write the failures there
in a syntax which can redrive the restore process.
2). After the first restore phase, redrive the restore using the failuresfile
as input.
3). Files which still fail should be stored in a 2nd failuresfile with a name
derived from the first. At this point the restore stops and the user must
take over.  If all that is needed is to kick a user or 2 off the server, then
the administrator can again redrive the failures with
'dsmc -retryfile=<filespec>'.

Editorial comment:  I have been an MVS systems programmer for 20+ years.  I
seen the growing pains of MVS and also HSM which I have administered for 15
years.  Now I am watching the growing pains of ADSM.  The ADSM developers
could make ADSM support large systems better and sooner if they had a
conference with their HSM neighbors about the steps they have been thru
to support large systems.


Bill Colwell
C. S. Draper Lab
Email: BColwell AT draper DOT com
Voice: 617-258-1550