catching up with sync logs

[Bron Gondwana]

On Tue, Jun 30, 2009 at 10:14:15AM +0200, Rudy Gevaert wrote:
> Hello,
> 
> Last Sunday we had some corruption on the lun of one of the replica's.  
>   After recovering it we got it back on line, but we are having some  
> backlog with the sync replication.
> 
> We have several sync log-* files that need to be processed.
> 
> While rolling replication is now busy on the normal log file. I'm  
> trying to feed an other sync client the other log files.  But it isn't  
> doing much.  Even on small log files.  E.g.:
> 
> read(8, "MAILBOX \"ugent.be!user.maarten^b"..., 4096) = 456
> time(NULL)                              = 1246349375
> read(8, "", 4096)                       = 0
> write(6, "\27\3\1\0000m\n  
> *\31\215|\303ke\331(\4p\236Q\21Z\313\341"..., 53) = 53
> time(NULL)                              = 1246349375
> read(6,
> 
> and then it waits... and waits.
> 
> Any tips and tricks on getting this back log fixed would be appreciated!

What's happening at the other end?  It looks to me like you're either
waiting on locks, or the process at the other end has died.

(worst case, some corruption isn't being handled correctly by sync_client,
and is causing protocol alignment issues, but I suspect locking or death)

Bron.