saving fr1 bitmap ?

Hi Peter, just a suggestion :

let's say an enbd-fr1 setup, for example a redundant Samba server

In normal operation :
A is enbd-client, with an fr1 array served by samba
B is enbd-server

- When B fails, everything is ok, A is using bitmap intelligence to avoid total reconstruction.
- When A fails, things are more complicated : If B mounts directly its local disk and serves it, 
when coming back to normal operation, A disk will be totally resync. There is a way to avoid it :
	1 - B has to mount its local disk via a degraded fr1 array,
	2 - when A comes back, it starts an enbd server, B starts a client, and add it to its fr1 		array, 
so it can use the bitmap for resync.
	3 - after resync, B stops everything, starts its server, A stops its client.
	4 - A and B starts normal operation with a clean array.

The problem is that these scenarii are quite difficult to implement cleanly via heartbeat scripts ( 
i did it, it works, but it is not really simple)

A simpler way to do that may be to allow bitmap created in step 1 on B to be saved to a temporary 
file when stopping array on B. This way, A will just have to download the bitmap file (scp, for 
example) and then start its array with this bitmap, allowing fast sync in any cases.
	

Is there already an user-space access to the bitmap, via /proc ?

This may imply some changes to mdadm tools too,  to save and load bitmap. Stop and Assemble modes 
are concerned. For example, an option like that  :


mdadm -S /dev/md0 --bitmap-file=/temp/md0
mdadm -A /dev/md0 /dev/sda /dev/nda --bitmap-file=/temp/md0


What do you think about it ?

Denis