[Gluster-devel] Confused on AFR, where does it happen client or server

[Brandon Lamb]

I have been reading through old mails for this list and the wiki and i
am confused on what machine would do the writes for an afr setup.

Say I have two "servers" (192.168.0.10, 192.168.0.11) with configs

-------------
volume locks
  type features/posix-locks
  subvolumes brick
end-volume

volume brick
  type storage/posix
  option directory /mnt/raid/gfs
end-volume

volume server
  type protocol/server
  subvolumes brick
  option transport-type tcp/server     # For TCP/IP transport
  option client-volume-filename /etc/glusterfs/glusterfs-client.vol
  option auth.ip.brick.allow 192.168.0*
  option auth.ip.brick-ns.allow 192.168.0*
end-volume
------------

and one client (192.168.0.20) with config

------------
### Add client feature and attach to remote subvolume of server1
volume brick1
  type protocol/client
  option transport-type tcp/client  # for TCP/IP transport
  option remote-host 192.168.0.10 # IP address of the remote brick
  option remote-subvolume brick     # name of the remote volume
end-volume

### Add client feature and attach to remote subvolume of brick1
volume brick2
  type protocol/client
  option transport-type tcp/client  # for TCP/IP transport
  option remote-host 192.168.0.11 # IP address of the remote brick
  option remote-subvolume brick     # name of the remote volume
end-volume

### Add AFR feature to brick1
volume afr1
  type cluster/afr
  subvolumes brick1 brick2
end-volume

### Add IO Threads
volume iothreads
  type performance/io-threads
  option thread-count 4
  option cache-size 32MB
  subvolumes afr1
end-volume

### Add IO Cache
volume io-cache
  type performance/io-cache
  option cache-size 32MB             # default is 32MB
  option page-size 1MB               #128KB is default option
  option priority *.php:3,*.htm:2,*:1 # default is '*:0'
  option force-revalidate-timeout 2  # default is 1
  subvolumes iothreads
end-volume

### Add writebehind feature
volume writebehind
  type performance/write-behind
  option aggregate-size 1MB
  option flush-behind on
  subvolumes io-cache
end-volume

### Add readahead feature
volume readahead
  type performance/read-ahead
  option page-size 256kB
  option page-count 16       # cache per file = (page-count x page-size)
  subvolumes writebehind
end-volume
------------

So when my client machine copies a 10 megabyte file, does the client
send this file to server1 and server2, or do the servers figure this
out?
Is there server side afr?

My question is along the lines of what happens if i have 10 servers
(data nodes) and 2 clients, how can i move the load of copying files
from clients to servers? or is this even possible?

I would think that in a mail cluster setup, i would not want my mail
servers having the extra load of handling the afr, i would think it
should send a write to one server and the server handles replicating?

Or am I totally not understanding something?