Daniel - An idea and questions - 1. Are you fsync()'ing the directories after you create them? Could you try that? 2. Are you sure you are only accessing Gluster via the Gluster mountpoint? 3. What version of Gluster? 4. What OS, version and file system are you using? 5. How many storage servers, what is your distribution model? (Replicate, pure distribute?) 6. How are you mounting Gluster? How many clients? 7. Can we get the exact error your script is getting? 8. Can you send the Gluster error log from one of the affected storage servers? Thanks, Craig --> Craig Carl Senior Systems Engineer Gluster From: "Daniel Goolsby" <danielgoolsby at gmail.com> To: gluster-users at gluster.org Sent: Thursday, October 21, 2010 3:28:10 AM Subject: caching. I seem to have some kind of caching issue. I have a process that will create hundreds of directories, then immediately spawn a parallel process across multiple nodes. The job ends of failing because some of the nodes cannot see the directories that the first process created. If I wait a few minutes then start the parallel process, it will succeed, which makes me believe some of the nodes just don't see the directories yet. Is there a way to disable caching (or if someone knows of something else I could check, I'd appreciate)? I saw there was an "option cache-timeout" but didn't get any better results setting that to 0. Thoughts? -- --daniel -- _______________________________________________ Gluster-users mailing list Gluster-users at gluster.org http://gluster.org/cgi-bin/mailman/listinfo/gluster-users
