[Info-vax] How do I avoid a full disk shadow copy when I shutdown/boot a node back into the cluster?

[Keith Parris]

On 4/4/2012 1:03 PM, Len Whitwer wrote:
> How do I avoid a full disk shadow copy when I shutdown/boot a node
> back into the cluster?

I'm wondering if it is a Full-Copy or a Full-Merge that you are 
experiencing?

> Have two rx3600's with a SAN on each running OVMS I64 ver 8.4 with
> clustering and Volume Shadowning.

You mention there's a SAN on each node. Is that two different SANs or 
did you mean that there is one SAN with all the disks accessible from 
both nodes?

If the SANs are separate, that would imply MSCP Serving is required for 
one node to access the disk on the other node's SAN. In this case, 
rebooting of the node would remove access to the shadowset members on 
that node's SAN for more than SHADOW_MBR_TMO seconds, causing the other 
node to remove that shadowset member from the shadowset, requiring a 
Full-Copy to resynchronize when the node reboots and access is restored 
to the disks on that SAN. If this is the case, then what you want is a 
Mini-Copy operation to resynchronize, and $DISMOUNT/POLICY=MINICOPY done 
from the node which will remain up will allow a Mini-Copy instead of a 
Full-Copy on the node's return. For this to occur automatically even in 
the event of an unplanned node outage, Google AMCVP, Automatic Mini-Copy 
on Volume Processing, to see how to set this up.

If there is one SAN to which both nodes have direct access, then it's 
most likely that the shadowset isn't getting properly dismounted and a 
Full-Merge is the result. If this is the case, then setting up a 
Mini-Merge policy should do the trick.

To clarify:

A Merge operation is required when a node leaves the cluster while it 
has a shadowset mounted.

A Copy operation is initiated when a $MOUNT command is invoked to add a 
shadowset member into a shadowset.