Another question about afr

[Albert Shih <Albert.Shih@xxxxxxxx>]

Hi

Me again with my afr question.

I want use afr for making my cluster node_down_proof.

My cluster is not for storage but for computation. It's for this reason if
a node is down we can allow the data on this node is not accessible (event
lost). 

But I want when one node is down the other node work fine.

For this reason I use afr with two replication (and now with Arnand Avati
it's working). But what's I don't understand is when I make one node (where
is the namespace) down (by using /etc/init.d/glusterfs-sserver stop)if I'm 
on one node I can access the data (not on the down node off-course), but I 
can make «ls» or sometime «cd». 

This my config


    volume node0-ns
      type protocol/client
      option transport-type tcp/client     # for TCP/IP transport
      option remote-host node0
      option transport-timeout 30
      option remote-subvolume name-space
    end-volume
    volume node1-ns
      type protocol/client
      option transport-type tcp/client     # for TCP/IP transport
      option remote-host node1
      option transport-timeout 30
      option remote-subvolume name-space
    end-volume
    
    #
    # Definition des namespace via afr
    #
  
    volume name-space
      type cluster/afr
      option replicate *:2
      subvolumes node0-ns node1-ns
    end-volume

    #
    # Definition du premier unify
    #

    volume unify
      type cluster/unify
      subvolumes node0 node1 node2 etc...
      option scheduler nufa
      option nufa.local-volume-name quadri11
      option nufa.limits.min-free-disk 10%
      option namespace name-space
    end-volume


Regards.

JAS

--
Albert SHIH
Observatoire de Paris Meudon
SIO batiment 15
Heure local/Local time:
Mar 18 déc 2007 00:44:34 CET