Thanks for the clarification, it's what I had figured but wanted to make sure. Have another question: is it possible to dynamically add AFR nodes (preferably without bringing down the server processes)? I added an entry for a non-existent server yesterday in the AFR configuration and it appeared to hang the process when I would attempt to access the client data. It may have been due to invalid network configuration on the server I was working on. I would like to know before I experiment much more (and waste time) if it's even possible. Daniel Maher wrote: > Justin Kolb wrote: >> If you have an AFR server and a process on the server writes or reads >> directly from the server's mount point (not using a glusterfs client >> connection) how does that affect replication? > > Interacting with the ? raw ? data is a generally a no-no. > Interactions should be done via the client or else they will not be > reflected on any other members of the storage group. > >> correctly, or if doing this negatively affects the system. If so is >> there a good way to help to avoid this type of direct access? > > There's nothing stopping you from running both a client and a server > process on a given machine. It's the same binary - just a different > invocation. > > > > -- > dan. > _______________________________________________ > Gluster-users mailing list > Gluster-users at gluster.org > http://gluster.org/cgi-bin/mailman/listinfo/gluster-users
