Red Hat Bugzilla – Bug 92187
Poor NFS server performance under load
Last modified: 2007-04-18 12:54:14 EDT
From Bugzilla Helper:
User-Agent: Mozilla/5.0 (Windows; U; Windows NT 5.1; en-US; rv:1.2.1) Gecko/20021130
Description of problem:
The server is a Dell PowerEdge 1650 2x1.13 GHz PIII.
I have a large RAID filesystem exported, which is connected to an Adaptec 39160
controller. I also have a second filesystem exported from the internal PERC 3di
RAID controller for testing.
Read/write performance from clients is mostly ok, until I access the filesystem
while logged into the server itself. For example, if I create a file using dd
on the server, and attempt to do the same from a client, the client gets nearly
zero IO until the server is finished creating the file. Also, the server
responds very slowly to terminal input while this is happening. Creating files
from two separate NFS clients sees good performance on both.
I have tried both NFS mounting the filesystem back to the server, and accessing
the filesystem directly from the server. Both result in the same extremely poor
performance for all NFS clients if a file access occurs on the server itself.
This happens for both filesystems described above.
I've upgraded the nfs-utils package to 1.0.3 and didn't see any change.
Version-Release number of selected component (if applicable):
Steps to Reproduce:
1. Export NFS filesystem
2. Attempt to write to the filesystem from both the server and an NFS client
3. Observe slow server respone and poor IO performance.
Joshua, Do you see slowdowns if you were to use the Redhat 2.4.18-xx kernels