[Gluster-users] simple AFR setup, one server crashes, entire cluster becomes unusable ?

[Daniel Maher]

Stas Oskin wrote:

> Based on my limited knowledge of GlusterFS, the most reliable and 
> recommended way (in wiki) is client-side AFR, where the clients aware of 
> the servers status, and replicate the files accordingly.

I've reviewed the AFR-related sections of the documentation on the wiki...
http://www.gluster.org/docs/index.php/GlusterFS_Translators_v1.3#Automatic_File_Replication_Translator_.28AFR.29
http://www.gluster.org/docs/index.php/Understanding_AFR_Translator

Nowhere in those sections is it stated, either directly or implicitly, 
that client-side AFR is more reliable than server-side AFR.  I'm not 
saying that the statement is incorrect, but rather that the 
documentation noted above doesn't seem to suggest that this is the case.

How, exactly, does relying on the clients to perform the AFR logic 
become more reliable than allowing the servers to do so ?  In either 
case, Gluster is responsible for all of the transactions, and for 
determining how to deal with node failure...

I am also curious about the network traffic with such a change.  In the 
current setup, the overhead of replication is restricted to two nodes - 
the servers.  Perhaps i mis-understand client-based AFR (which is 
entirely possible!), but i suspect that my replication overhead would 
increase for each client, since each client would send writes to both 
servers.  Currently this isn't a problem, but as the number of clients 
increases, so would the overhead - correct ?  We intended to double the 
number of servers as well (remote site) - wouldn't this in turn double 
the replication overhead for each client ?  This would get out of hand 
fairly quickly...

Don't get me wrong, i am more than happy to try a client-based AFR 
config if it truly is superior ; however as of right now i don't know 
how or why this would be the case.

Thank you all for your continued suggestions and discourse.


-- 
Daniel Maher <dma+gluster AT witbe DOT net>