[svlug] RH-EL, slow NFS or slow network

Roland Krause rokrau at yahoo.com
Mon Jul 18 16:34:30 PDT 2005

Hi all, 
I've got a problem with a small-to-medium size network and a NFS file
server. The NFS file server is just too slow and I would like to find
out what the problem is. Any suggestion to good network monitoring
tools that let me investigate this problem are highly welcome. 

The NFS server is a DELL PowerEdge with 1.8TByte DELL PowerVault SCSI
arrays, it's running RH-EL3-U4 and has channel bonded Dual Gig
connection to the switch. 

The switches are Nortel Baystack, there are four of them in stacked

There are about 50 Unix clients, about 40 running Linux RH9 and SuSE-92
(32 and 64bit), 10 boxes are running Solaris-8. There are also  about
40 Windows XP-Pro that don't do NFS. Everything runs on the same
subnet. The NFS server exports Unix directories via SMB. 

There is some junk on the network like old 10Mbit/s print servers, the
cards are a mix of 100Mbit and 1Gbit cards mostly Broadcoms but really
a mix. 

One symptom of the problem is that, whenever someone writes a large
file (20-30MBytes) to the file server, there is a very noticeable
response lag for everybody else. E.g., bash completion stalls for about
one or two seconds. 

Another symptom is that when copying (1GByte) files between systems,
using RAM disks, we see a maximum of about 150Mbit/s throughput. 

Finally, the weirdest problem: Some user changes a file on his NFS
mounted home directory and the changes are not apparent on other
clients for many minutes! 

Where do we start looking? Any suggestions are highly welcome. 

Best regards,

PS: I have had a lengthy discussion with my "IT-department" about
moving the Unix network on its own subnet. This was denied because
there would be no advantage in doing so - as to my Windows network
experts :-) - I need some good counter arguments for this. 

Start your day with Yahoo! - make it your home page 

More information about the svlug mailing list