UNIX for Dummies Questions & Answers

Hi All,
i'm experiencing some strange behavior (at least for me) of my NFS server.

In few words I have a server that exposes a directory via NFS and a client that mount it.
I tested a simple copy from the client to the exposed directory of a file of 2GB and i see that on the server the file is created pretty soon, then no byte is written for a while (checking with ls -la), then the size increases from 0 to something like 8MB, then after a while to 12MB and so on....clearly with some buffer since it doesn't increase "linearly".

This is how the directory is exported:

Code:

[root@server ~]# exportfs -v
/nfs/dir my-client(rw,async,wdelay,root_squash,no_subtree_check,fsid=0,anonuid=65534,anongid=65534)

And in the fstab client file i have:

Code:

server:/ /nfs/dir nfs4 rw,rsize=32768,wsize=32768,intr 0  0

Is there any option to remove that buffered behaviour ?

Thanks in advance
Ste

Well, the only things really fair to complain about is the after state and speed. If the size is wrong after the copy, then you have a fair gripe

I wrote an mmap() process on NFS file space and the changes did not propagate unless I did an ls -l on the dir to force an update to the local view of the remote inodes..

I wrote another process that effectively did a tail -f of an nfs mounted file on os/390 in C, for near real time monitoring with 100% logging, and to make sure it did not stall, just did close, open, seek and if no new data, poll(0,0,1) (millisecond sleep to avoid hogging CPU); this worked great.