We have been having some problems with the linux NFS client talking to
an IRIX NFS server.
Here is the scenario. Our developers have SGI boxes on their desktops
and they are editing their code on those boxes. They also have a Linux
RH5.2 box running 2.2.1 kernel. When they are done making changes to
their code they save their files and then run make in the Linux box
window. The Linux box might see the changes to the file in about 30-60
seconds after they save the file. Our developers are saving their files
to an SGI NFS server.
I tried the following test.
SGI client box A
SGI client box B
On SGI client A, I created a file and saved it to SGI server. I then
tried to cat the file on SGI client B and on Linux client. SGI client B
was almost instantaneous, and I had to wait for several seconds before
I could cat the file on Linux client
Is there any way to speed this up? Instead of jacking up the attribute
caching min and max is there a way to force the Linux client to flush
the attribute cache and re-read the directory contents?
Also on a related topic why do we see very little latency on SGI boxes
when talking to the same server when all mount parameters (attribute
caching, etc) are set the same way?