[Info-vax] more on minicopy

[Phillip Helbig---undress to reply]

I'm even more anxious to upgrade not just to 8.3 but to 8.4 because of 
the HBMM multiuse bitmaps.  Glad that VMS had already implemented an 
idea I thought of.  :-)

It seems to me that intentional dismounts of members of shadow sets are 
rare compared to unplanned node crashes and thus I find it strange that 
this functionality wasn't in the original minicopy implementation, 
especially since HBMM was around.  Anyone know the details?

Apparently DISMOUNT/POLICY=MINICOPY to dismount a shadow-set member
creates the bitmap on the dismounting node only, and if it goes down,
the bitmap is gone and the only alternative is a full copy.  However, it 
appears that things are more robust as long as no nodes unexpectedly 
fail.  For example, one can issue DISMOUNT/POLICY=MINICOPY on one node, 
issue a MOUNT/POLICY=MINICOPY on a second node and the minicopy can be 
done by a third node.

When a node goes down unexpectedly, taking shadow-set members with it, 
then when the node reboots they could come back with a minicopy as long 
as there is a minicopy bitmap anywhere in the cluster, as far as I can 
tell.  The problem before 8.4 is having such a bitmap before the member 
unexpectedly disappears; even if one creates one by intentionally
mounting less than all shadow members, as soon as one enters the shadow 
set it uses up the minicopy bitmap which then gets deleted.