Jean Figarella wrote: > My old nfs server was an old dual Xeon machine running debian. > Everything worked fine. Then I switched over to a new Penguin Computer > Altus 2600 (if Im not mistaken) with two Dual-Core AMD Opteron(tm) > Processor 2210 and 4GB of ram. CentOS4.4 is running on it. > > Now here is the thing, this nfs server is exporting "/home" and all > workstations mount the /home dir via fstab. Ever since I switched to > this new server running CentOS4.4 everyone on the network gets 5-20 > second freezes sporadically trough the day. In other words if I am > browsing the net using Firefox, because all the user's options and cache > are stored on ~/.firefox, it freezes for the 5 to 20 seconds at a time. > > After running some test, this happens when big chunks of file are being > copied via nfs to the server. So as an example if user1 has a 700Mb iso > image in /tmp and he decides to copy this to his home dir (for whatever > reasons, this is not important), every user in the network will > experience 5 to 20 seconds freezes while the file is being copied. Now > if the same file is being scp-ed to the same user's home dir on the > server, the freezes do not happen. > > Now I have tried many different nfs server options and other nfs client > options. I also set up a test nfs server with centos4 and 5 where I am > the only user, and while copying big chunks of date, nfs freezes (hangs) > too. > > Here is my /etc/exports: > /home *domain.com(rw,root_squash,sync,no_subtree_check,no_wdelay) > > Here is how client workstations mount the export in the fstab > fshome:/home /home/ nfs rsize=8192,wsize=8192,intr 0 0 > > Am I mis-configuring the nfs server of clients? Is there anything I am > not doing right? > If anybody has any suggestions, I would appreciate it very much. > > Thanks, > Jean Figarella. > _______________________________________________ > CentOS mailing list > CentOS at centos.org > http://lists.centos.org/mailman/listinfo/centos I think I found the issue here. Its the sync option in the server. I am running it with async instead and so far so go. But still any other suggestions are welcome. Jean