On Mon, Apr 02, 2001 at 08:56:08PM -0500, Avi Schwartz wrote:
I had no problem reading and writing multiple large files, 21 MB and 26 MB files to and from NFS mounted share. Both machines run Linux and the ^^^^^^^^^^^^^^^^^^^^^^^
This is the difference in our setups. I am using shares exported from Solaris 2.7 system. Thanks Avi for the test!
partitions are ReiserFS.
Avi
Actually, I found a message on Linux Kernel mailing list pointing to a bug recorded in Red Hat bug tracking system http://bugzilla.redhat.com/bugzilla/show_bug.cgi?id=30944 but it refers to NFS server on SGI Irix 6.5.x and symptoms are different. The error is a missing directory when reading from NFS share. Can we expect an NFS patch from SuSE for 2.4.2/2.4.3 kernel? Thanks, -Kastus
--On Monday, April 02, 2001 18:28:32 -0700 "Konstantin (Kastus) Shchuka"
wrote: Avi,
You are right about fixing lockd under 2.4.2 but you are wrong stating that 2.4.2 is ready for prime time.
NFS client in 2.4.2 has serious problems.
I mount an NFS exported share from Solaris 2.7 onto my 7.1 system. Everything is fine when reading from NFS. When I try to copy a big (20+ Megs) file to NFS the cp command hangs and then spits input/output error message. Nothing gets copied.
I reboot my Linux box into 2.2.18 kernel and repeat the same cp command. File gets copied.
That means 2.4.2 kernel is ready for prime time with NFS for read-only. If that's the definition of prime time, let it be so...
Could anybody on the list repeat this test, writing a big file to nfs-mounted disk using SuSE stock kernel 2.4.2?
Thanks,
-Kastus
On Mon, Apr 02, 2001 at 07:37:16PM -0500, Avi Schwartz wrote:
Unlike what other are saying, 2.4.2. is ready for prime time. I am attaching a mail message I sent this list last week regarding this problem. Is show how to fix it. The problem is with the script that starts NFS under the 2.4.X kernels:
When starting up the nfsserver under Kernel 2.4.2 I get the error message lockdsvc : invalid argument. Searching the Internet (thanks Google!), I found some messages in the linux-kernel mailing list that there is no need to start rpc.lockd from the init scripts any more when using rhe 2.4.X kernels as it is loaded automatically on mount. I commented out these line from /etc/rc.d/nfsserver:
checkproc -n lockd || \ /usr/sbin/rpc.lockd
and restarted the NFS server. The error is gone and so far everything seems to work.
Avi
--On Monday, April 02, 2001 15:20:33 -0700 "Konstantin (Kastus) Shchuka"
wrote: Hi,
<rant mode>
I lost whole Friday trying to upgrade my 7.0 box at work to 7.1. Finally I gave up understanding that the only way to upgrade is do a clean install.
I would appreciate if that were written in BIG letters somewhere on the box.
I did a clean install, chose kernel 2.4 as a default, system booted.
Initially I didn't pay attention to the message:
lockdsvc: Invalid argument exit status of /usr/sbin/rpc.lockd: 1
from rcnfs script.
Later I discovered that I cannot copy a big (19Megs) file to a nfs-mounted disk from Solaris. There were no problems earlier when I was using SuSE 7.0.
I can copy a small file, I can create directories, but as soon as I try to copy something big, I receive input/output error.
I installed then 2.4.2 kernel form SuSE, and nfsutils-0.2.1-20. Nothing changed, same hanging and error while copying big file over nfs.
Is this a feature of 7.1 with 2.4.2 kernel that NFS should not work?
-Kastus