Gluster users - Jun 2009 - Poor Performance on a replicate client

Hello,

I am just playing around with clusterfs version 2.0.1 on two openSuSE 11.1 
systems. I am using the replicator translator. My client and server config are 
below this mail. i have two servers and one of the servers acts also as 
client. Basically this setup works. 

The problem is that performace writing to the replicated volume is really 
poor. I tried to copy /etc  (running: find /etc/ |cpio -pdmv /replicatedvol) 
to the replicating volume and this took 4 minutes. In contrast doing the same 
locally on the filesystem takes only 4 seconds. The complete size of /etc is 
95MB. So the write performance was about 0,395 MB/sec for glusterfs. Is my 
config wrong it seems to me that this performance is very  poor?  Both servers 
are connected by a 100 Mbit/sec switched network that is not buisy. 

Another question is about recovery. I killed glusterfd on one of my two 
servers (say on server "B")  then copied data on the replicated
directory from
my client. Next I started the server "B" again I had killed before. My
expectation was, that it finds out that server "A" has data in its
replicated
directory that are not yet on "B".  But nothing happened. Both servers
came
only in sync when I accessed the replicated filesystem via the client eg by 
doing an ls . Is this the only way to sync to replicating servers? 

Here are my current testing configs. The Server config is running on two 
openSuSE11.1 machines, the client config is running on one of the two 
machines:

Server:
volume posix
  type storage/posix
  option directory /cluster
end-volume

volume locks
  type features/locks
  subvolumes posix
end-volume

volume brick
  type performance/io-threads
  option thread-count 8
  subvolumes locks
end-volume

volume server
  type protocol/server
  option transport-type tcp
  option auth.addr.brick.allow *
  subvolumes brick
end-volume


Client:
volume remote1
  type protocol/client
  option transport-type tcp
  option remote-host rzinstal2.uni-koblenz.de
  option remote-subvolume brick
end-volume

volume remote2
  type protocol/client
  option transport-type tcp
  option remote-host bliss.uni-koblenz.de
  option remote-subvolume brick
end-volume

volume replicate
  type cluster/replicate
  subvolumes remote1 remote2
  option  metadata-self-heal  on
  option  data-self-heal on
  option  entry-self-heal  on

end-volume

volume writebehind
  type performance/write-behind
  option window-size 1MB
  subvolumes replicate
end-volume

volume replicateclient
  type performance/io-cache
  option cache-size 512MB
  subvolumes writebehind
end-volume


-- 
Rainer Krienke, Uni Koblenz, Rechenzentrum, A22, Universitaetsstrasse  1
56070 Koblenz, http://www.uni-koblenz.de/~krienke, Tel: +49261287 1312
PGP: http://www.uni-koblenz.de/~krienke/mypgp.html,Fax: +49261287 1001312