I didn't mean to criticize GlusterFS at all, just taking Daniel's word how > GlusterFS works. I'm quite a noob at GlusterFS in particular and large > storage solutions in general. > > If it is true that the files will be pushed back and forth by just updating > one byte in them it will be an issue however if that's not the case then I > can carry on with my first intent which is to use GlusterFS for storing and > accessing indexed data. > Under normal conditions, if you write 1 byte to a file, only 1 byte is written to all servers. But, if during writing 1 byte, one of the servers was down, then 1 byte is written to the reamining server(s). when the server which was down comes back up, and the file is accessed again, then the entire file content from the updated server is copied to the lagged server (this is the self-heal process). We are working on improving the healing efficiency for such situations (minor changes on large files) avati
