[Info-vax] ANOTHER minicopy question

[Phillip Helbig---undress to reply]

In article <op.wqcqkxo8i0cpeq at kparris4.americas.hpqcorp.net>, "Keith
Parris" <keith.parris at hp.com> writes: 

> On Sat, 29 Dec 2012 10:19:21 -0700, Phillip Helbig---undress to reply  
> <helbig at astro.multiclothesvax.de> wrote:
> > I'm wondering how to set up MINICOPY on a system-disk shadow set,
> > considering that it is not explicitly mounted during the startup
> > sequence and is mounted by the booting node before any other node can
> > mount it.  HELP says:
> 
> If you try to reboot a node from its local copy of the system disk  
> shadowset, you're doomed, because when the node gets far enough along in  
> boot to discover that it has booted from a now-out-of-date copy of the  
> system disk shadowset, it will do a bugcheck of some sort regarding  
> inconsistent shadowset state and leave the cluster.

What was confusing me is that this HELP text assumes that not all 
members are mounted.

> You might consider a combination of two things:
> 1) Set up Automatic Mini-Copy on Volume Processing (AMCVP). This requires  
> OpenVMS version >=8.3 (plus Shadowing patches to fix the initial bugs an  
> AMCVP). With AMCVP, you set up HBMM bitmaps in your policy as Multi-use,  
> which allows an HBMM bitmap to automatically be converted to a Mini-Copy  
> write bitmap at the time of the loss of a shadowset member.

Right; another reason to upgrade to 8.4.  I was assuming that this was 
how the old minicopy worked.

> 2) Boot each node by default as a satellite node of the other 2 nodes.  
> This allows the booting node to boot from the present 2-member system disk  
> shadowset instead of its out-of-date local copy, and then once it's up you  
> could add its member to the system disk shadowset with a Mini-Copy  
> operation.

I have never thought of that.  However, I should mention that both 
members have direct connections to only one node, the node which boots 
from them.  I was wondering how to set things up so that if a disk went 
temporarily offline, but the node stayed up because the other member was 
there, it could get a minicopy when coming back.