[Gluster-users] architecture question

ruben malchow ruben.malchow at googlemail.com
Mon Feb 18 18:33:17 UTC 2013

hi everyone,

i am having a few difficulties understanding some of the concepts - especially (i think) because i am not always sure wether i am reading 3.2 docs or 3.3 …

the situation is this: we are doing an evaluation for a rather large system that will be used to store video data. we will need around 500TByte of storage and a sustained
throughput of somewhere north of 3GByte/s. we are in the process of setting up an initial, scaled-down version of this.

the current setup is this: 

we have 10 rather stupid storage nodes with 4 bricks and a single 1GBE connection, and another, more powerful machine with a 10GBE interface on the same network as the other nodes and another 10GBE interface on the same network as the clients. 

the idea was to have gluster stripe & replicate across these 2x (replicate) 20x (stripe) bricks, then have a client running on the big machine that exports the volume to the actual client machines. 

i have included the current config below, but i think i am missing something fundamentally … ? and, maybe i am getting the terminology and lines dividing cluster, server and client wrong?

any help would be appreciated!

thank you!


# local bricks, 4 of these:
volume brick1
 type storage/posix
 option directory /media/brick1


volume brickN
 type storage/posix
 option directory /media/brickN

# 40 of these:
volume gv0-brick-$HOST_NUMBER-$BRICK_NUMBER
   type protocol/client
   option remote-host 10.25.178.$HOST_IP
   option remote-subvolume $BRICK_ON_HOST  # (i.e. brick1)
   option transport-type tcp

#  stripe 1
volume gv0-stripe-01
 type cluster/stripe
 option block-size 1MB
 subvolumes gv0-brick-01-01 gv0-brick-02-01 […]

#  stripe 1
volume gv0-stripe-02
 type cluster/stripe
 option block-size 1MB
 subvolumes gv0-brick-05-01 gv0-brick-06-01 […]

#  replicate
volume gv0
   type cluster/replicate
   option background-self-heal-count 0
   option metadata-self-heal on
   option data-self-heal on
   option entry-self-heal on
   option self-heal-daemon on
   option iam-self-heal-daemon yes
   subvolumes gv0-stripe-01 gv0-stripe-02

volume server
 type protocol/server
 option transport-type tcp/server
 option auth.addr.gv0.allow 10.25.178.*,
 subvolumes gv0

More information about the Gluster-users mailing list