On Wed, 24 Jun 2009 09:45:54 +0100, Vidar Hokstad <vidar@xxxxxxxxxxx> wrote: > Hi, > I have a cluster set up with 2.0.1, and when under heavy load I keep > getting > files that are getting cut off in the middle (cat [problematic file] on the > client will give different/shorter output than doing the same on the > server) > When I unmount and mount the client volume again it works fine for a while, > so I'm suspecting either the io cache or readahead translators. Is this a > known problem? That sounds vaguely familiar. Didn't somebody mention a patch for this recently? > If it is a known problem is it fixed in 2.0.2 (which don't compile on two > of > the machines in the cluster for me for some reason, but the two problematic > ones are really old and messily set up so I'm assuming that's a problem > with > their setup - it's kept me from testing 2.0.2 so far, though) or in git? Compile on a clean build machine and just install the package on the ones with a broken build environment? > If it isn't a known problem I'll gather up some more info and do some more > work to try to get a newer version working so I can see if it happens on > 2.0.2 or the git version as well. I suspect you'll find that glusterfs is still undergoing a sufficiently rapid development and stabilization process that staying a few versions back is very counterproductive. In the majority of releases more gets fixed than broken, so I wouldn't even bother reporting an issue with an older version unless you have verified that it is still broken in the latest release (unless it's a regression, i.e. something that used to work but got broken in a newer version). Gordan
