Hi Amar, Thanks for your answers, my replies follow inline. On 4/1/12 2:15 AM, Amar Tumballi wrote: > On 03/22/2012 08:51 AM, Daniel Pereira wrote: >> Hi everyone, >> >> Hope anyone can shed some light on some strange behavior we've just >> having with Gluster. >> Last night, while using gluster3.3b2, we came up with some errors, after >> running tests for several days. >> > > Hi Daniel, > > Can you consider testing 3.3.0qa32+ versions untill 3.3.0beta3 is made > available? Because 3.3.0beta2 was pretty old and we have made > significant changes/bug fixes after the beta2 release. We will consider it, but we noticed the error happened sporadically when using stable client versions (3.2.5) with 3.3b2 servers. We changed the clients to use the same 3.3b2 version and the problems seem to have disappeared. >> We're running the test in a striped 4 disk volume: >> # gluster volume info data >> Volume Name: data >> Type: Stripe > > Considering this is a 'stripe' volume, what is the backend filesystem > you are using? Because with XFS there seems to be some issues with > data block pre-allocation algorithm, and glusterfs's logic of leaving > holes for striped blocks. We are evaluating better ways of solving the > stripe issue, till then, I recommend ext4 or other filesystems for you > as backend. We are already using ext4 in all the volumes. Thanks for your time! I really appreciate having a reply from the team. Daniel
