Jeroen Koekkoek wrote:
I configured glusterfs to do afr on the server side e.g. the client only connects to the localhost and the glusterfs daemon listening on the localhost replicates the files over the network. Since I need file locking, I used posix-lock/translator volumes and let the afr volume use the posix-lock volumes on both hosts as subvolumes. That however doesn't work. I can still lock the same file on both hosts at the same time.
I never got this far - glusterfsd core dumps on me when I wrap the base volume with the posix-lock translator. :-(
After doing some testing I found out that if I configure the client to do afr and use the posix-lock volumes as subvolumes in the client, fcntl locking does work.
So you export the raw storage/posix volume from the server as tcp/server and wrap it in posix-locks on the client after importing it via the tcp/client? Interesting, I was told to apply the wrapping directly around the storage/posix brick.
I'd like to know if more people ran into the same problem and if anybody has a solution. I'd also like to know if there are downsides to my current configuration.
I'm still waiting for Avati to have a look at the dumped core on one of my servers.
Gordan
