[Pgcluster-general] Desynchro during recovery processing

[Philippe]

A little question about recovery processing 

I'm doing some test with:

1 LB
2 DB node
1 replication server

The two DB node are up and ready
Whe begin "insert" in a loop
Whe shutdown DB node n°2
Whe ending the loop of "insert"
We starting DB node n°2 in recovery mode
Recovery OK

The recovery is finish and the two db are identical..

But if there is writing during the recovery ( it's possible because 
replication server is up and db node n°1 is up )
So the DB node n°1 stay in read-write


The two DB node are up and ready
Whe begin "insert" in a loop
Whe shutdown DB node n°2
We starting DB node n°2 in recovery mode
Recovery OK
The "insert" loop continue for some query ..
Whe ending the loop of "insert"
Whe restart DB node n°2 in normal mode ( or not )

Maybe we must leave the DB node n°2 in recovery mode all time ? or restart in 
standard mode ? ( not important at this point )

The recovery was ok but the two database are different because of "insert" 
continue writing during the recovery ..

Is there a solution for queuing "insert" during rsync recovery ? so the queue 
can be replayed into DB node n°2 ?
Maybe it is made by the replication server ?

Thanks in advance,

Philippe