With millions of files on a system this is a HUGE overhead. Running the getfattr command just for the mismatched files and using this as a source for triggering self-heal/sync might be a better option and less costly... but you might see files in the process of being sync'd as well. I'd only trigger a repair after a certain time has passed and nothing has happened. I'm still puzzled what caused the mismatch and why it didn't get repaired on it's own? Any ideas? Best, Martin Am 28.04.2011 16:48, schrieb Whit Blauvelt: > On Thu, Apr 28, 2011 at 04:16:51PM +0200, Martin Schenker wrote: > > >> After triggering manually with "touch" using the right *CLIENT* >> mount points, the self-heal/sync function worked fine. I was using >> the server mounts before, as shown by the getfattr output. Not >> good... >> >> Now the question remains WHY the Gluster system didn't do anything >> on it's own? Is this a "healthy" situation and we shouldn't worry? >> > Would it be good practice to regularly run a script to trigger any > self-healing that might be necessary - or to test if necessary (how?) and > then run on that condition? It would be easy, for instance, to use Python's > os.walk function to run through and touch - or whatever - every file in the > space. That adds a non-trivial load to a system, but for systems with load > to spare, would running that say every hour, or every day, but good > practice? > > Whit >
