Skip to content
Prev 207340 / 398506 Next

ff package: ff objects don't reload completely on NFS drives from a different machine

Hi,

this could be due to how NFS works.  Note that there can be up to a 30
second delay before other hosts on the same file system see the
updates that was flushed by one machine.  You basically cannot treat
files on an shared NFS file system as if you are working on a single
machine.  You have to add some higher protection if your data sources
should be shared ...and that is not an easy problem if you want it to
be bullet proof.  You need to use a semaphore/mutex or other ways to
communicate when files are updated/flushed/read etc.  I'm still
looking for a such a mechanism done over a file system that is bullet
proof (without having to relying on a central server).

My $.02

/Henrik
On Sat, Jan 23, 2010 at 12:02 PM, Hao Cen <hcen at andrew.cmu.edu> wrote: