Description of problem: At thursday morning, nfs suddenly became *very* slow. So slow that loging in to the boxes took 3 minutes instead of 5 secounds - this problem has been windows-only so far (luckily i don't administer thos e boxes). After some digging, if found that the problem probably was do to slow responce from stat() - ls (which is an alias to ls --colours) tok 10 secounds in /home, while /bin/ls tok 0.01 secound. The machines mount their /home over version 2 nfs from a debian box, My out-of-date laptop do not have any problems, neither did the boxes have before the upgrade. It seems like transfer rate is okay, but opening/closing a file takes _a while_. Version-Release number of selected component (if applicable): nfs-utils 1.06-22.i386 How reproducible: Steps to Reproduce: 1. 2. 3. Actual results: Expected results: Additional info:
BTW how do i rollback (ie. where do i find a functional RPM? CD's?) and prevent yum from re-installing the faulty RPM? i use "nightly yum update". Guess i shouldn't...
Would it be possible to get a ethereal trace and the output of nfsstat -rc?
sure, but i have to downgrade most of the boxes asap. How do i isolate the output to NFS only?
Try the capture string of "port 2049"
Forget it. My mistake. The problem was not NFS, it was LDAP. Which matches with the libuser libary was just updated. Sorry!