Wondering if anyone else has seen or heard of something like this, got some users starting to complain about it.
Issue only effecting (it seems) sles9 hosts, more likely 2.6 kernel is the difference, not sles nor version 9. There's always been a delay in seeing changes to files from different hosts to a path in NFS. i.e. if I write to a file in NFS from host A, it'll sometimes take up to a minute (I've seen about a minute) for host B to see that change. I know this is due to the write cache on host A taking some time to flush, then host B taking time to re-read the data from the filer instead of it's read cache. What some users are seeing now though, is it taking hours, if not forever (until the writing app is exited) for host B to see the updates. I can't even think of a way to force a box to do this, let alone why it would happen on it's own. Maybe there's a know bug in caching I haven't read of? I searchd google a ton and didn't see anything like this. kernel 2.6.5-7.147 btw.. the -7.147 being suse's patch level. My brute force thought is to disable the NFS write cache alltogether.. but I'm sure that would entail a huge performance hit. -- Mike Marion-Unix SysAdmin/Staff Engineer-http://www.miguelito.org Marge: "Homer, sitting that close to the TV can't be good for you." Homer: "Talking while the TV's on can't be good for you!" ==> Simpsons -- [email protected] http://www.kernel-panic.org/cgi-bin/mailman/listinfo/kplug-list
