Emmanuel Dreyfus <manu@xxxxxxxxxx> wrote: > It seems much faster. Here are the result so far. I will post an update later. It completed my tar extracction test with decent performances. I then tired a replicated volume, which is also fine so far. My testbed where the problem exists has 4 bricks (replicated and striped), but I know suspect there is something wrong with the volume: I have been reusing it for testing for a while, and glusterfs crashed a lot on it. I am almost certain that the problem will vanish if I start from scratch. But perhaps you would be interested in exploring further why that glusterfs volume went into a poor performance mode? -- Emmanuel Dreyfus http://hcpnet.free.fr/pubz manu@xxxxxxxxxx
