We are having a weird issue where file creation/removal from a NFS client takes a very long time to propagate out to the other client(s).
Mount options we have on the clients are:
defaults,rsize=32768,wsize=32768,intr,noatime,cto
The export has:
*(rw,sync,no_root_squash,no_wdelay)
We verify this by on one client doing:
watch -n0.1 stat foofile
And on another client we touch foofile (or rm it). Modifications take 1+ seconds to propagate out.
cto and no_wdelay are options we've just recently added to see if they solve the issue (they dont). What else can we look at?