I seem to have found the problem. My client machines that were crashing each had 32GB RAM. During some previous testing, I had set a kernel parameter (mem=4096) to limit the amount of system memory in my grub.conffile and had forgotten about it. It appears that having this set somehow triggers the system crashes when writing large files via NFS.
Is this a bug, or does anyone know if this behavior should have been expected?
Thanks, Tom
On 1/17/07, bigendian@gmail.com bigendian@gmail.com wrote:
I have an unusual scenario where a new 64-bit multi-core Opteron Centos 4.4 client crashes when writing a large file on a 32-bit NFS server. The problem happens on the same client with either Centos 4.4 or SuSE 9.332-bit NFS servers. Additionally the problem is reproducible with another identical 64-bit Centos 4.4 client. I'm not certain that the problem is tied to a 32-bit (as opposed to 64-bit) NFS server, but it doesn't happen on a local file system.
The crash appears to be some sort of kernel panic, but I don't have the details as physical access to the server isn't easy. The crashes seem to happen when creating files over 2GB in size, but there doesn' seem to be a fixed size that triggers the problem. Sometimes I can get file sizes as high as 3GB before the crash, but it's ususally at the 2.3GB to 2.6GBrange that it happens.
Are there known issues with Centos 4.4 and NFS that I might be running into?
Thanks, Tom