Interesting! Since you are getting (relatively) high i/o wait states, that means you likely have enough nfs daemons on the server to process client requests, but nfs server is just too slow to process the client's i/o and it gets backed up.
You mention editing/compiling fortran programs; presumably small ones, so that isn't going to do much on the home directories. Firefox, however, can be brutal. Move the whole profile (copy+symlink is easiest), or maybe start with just moving the cache to the local workstation's disk (/var/tmp/?) or /dev/shm/firefox-$USER to keep it disposable. The cache directory can be set in firefox under the URL about:config -- search for "cache" and note the browser.cache.disk.parent_directory option. Good luck! PS - the "c" command in "top" will show the whole process name, not just a snippet. On 11/29/2012 01:49 PM, David Fitzgerald wrote:
I also ran pstree and I will put that output below, but I think I may be barking up the wrong tree. While some of my clients were freezing up, I saw that my NFS server was getting very high 'top' loads. Fortunately I have sysstat running on the server and after class 'sar -u' showed that %iowait went from less than 1 before class to a high of 53 after class began, and stayed high until class ended. Here is the relevant 'chunk' of the sar -u output: 05:20:01 PM all 0.03 0.00 0.07 0.17 0.00 99.73 05:30:01 PM all 0.03 0.00 0.03 0.11 0.00 99.83 05:40:01 PM all 0.18 0.00 0.50 1.88 0.00 97.44 05:50:01 PM all 0.16 0.00 1.12 6.93 0.00 91.78 06:00:01 PM all 0.73 0.00 5.23 32.61 0.00 61.43 06:10:01 PM all 0.77 0.00 6.55 53.67 0.00 39.01 06:20:01 PM all 0.13 0.00 4.81 27.81 0.00 67.25 06:30:01 PM all 0.13 0.00 6.69 21.71 0.00 71.47 06:40:01 PM all 0.11 0.00 3.47 33.34 0.00 63.08 06:50:01 PM all 0.11 0.00 3.20 31.02 0.00 65.67 07:00:01 PM all 0.24 0.00 3.93 30.79 0.00 65.05 07:10:01 PM all 0.16 0.00 3.63 20.51 0.00 75.71 07:20:01 PM all 0.18 0.00 5.23 1.45 0.00 93.13 07:30:01 PM all 0.10 0.00 5.72 0.70 0.00 93.48 Average: all 0.06 0.01 0.46 2.13 0.00 97.34 The NFS server is a virtual machine in running ESXI 4.1 and VMware tools IS installed. Could this be slow disk access, and thus a VMware misconfiguration? I hate to admit it, but I am at a loss.
