[Gluster-devel] AFR self heal and moved directories

[Norbert Paschedag]

Hi,

I'm having a problem with AFR and self heal after one of two servers 
restarts. If a directory is moved during the absence of the server,
the directory tree will not heal and the presented tree is a union of 
'before' and 'after'. The 'before' state (directories only) is even 
copied to the server that didn't restart.
Yes, I did all kinds of open() on files and directories.

The shown behaviour is very likely not the intended one. Can any of the 
developers please explain how directories are handled when repairing ?

Using 1.3.1, vol files below.

Regards,
  Norbert

# server 1 
volume brick-ds
    type storage/posix
    option directory /gexport/1
end-volume

volume server
    type protocol/server
    option transport-type tcp/server
    option bind-address 127.0.0.1
    option listen-port 8881
    subvolumes brick-ds
    option auth.ip.brick-ds.allow *
end-volume

# server2
volume brick-ds
    type storage/posix
    option directory /gexport/2
end-volume

volume server
    type protocol/server
    option transport-type tcp/server
    option bind-address 127.0.0.1
    option listen-port 8882
    subvolumes brick-ds
    option auth.ip.brick-ds.allow *
end-volume

# client
volume brick-ds-1
     type protocol/client
     option transport-timeout 4
     option transport-type tcp/client
     option remote-host 127.0.0.1
     option remote-port 8881
     option remote-subvolume brick-ds
end-volume

volume brick-ds-2
      type protocol/client
      option transport-timeout 4
      option transport-type tcp/client
      option remote-host 127.0.0.1
      option remote-port 8882
      option remote-subvolume brick-ds
end-volume

volume brick-ds-afr
      type cluster/afr
      subvolumes brick-ds-1 brick-ds-2
      option self-heal yes
      option replicate *:2
end-volume