[opensuse] NFS misconfig problem w/files >2GB
I'm exporting some files (>10GB files) via NFS from a 2.6.23.1-kernel, SuSE system, and mounting on another (same kernel). All of my files >2GB (not exactly sure if break point is at 2 or 4GB's, all the large files I look at on the source are over 20G) give stat errors over NFS. Both kernels have nfs v4 as well as v3 compiled in. I've heard v2 has a 2GB limit, but v3 should be fine. Have been playing with tags 'nfsvers=[3 or 4]' in the /etc/exports file, but exports doesn't recognize nfsver=4, and gives out of range error message on nfsver=3 saying allowed values are Min=1, Max=2. !? I'm not sure if it is using V3 or 4 (am trying to use 'tcp' tag as well, but not using the tcp isn't the prob, as that was the default before I started looking at this problem. I don't have problems over CIFS from a windows box, but so far, limited benchmarks show better linux-to-linux perf over NFS (assuming I stay under 2GB files, but that's sorta limiting...). Underlying filesystem on the server box is XFS -- been serving Win clients for years, so haven't had alot of experience using NFS, recently, since >2G files became fairly common place. Any ideas how I could be serving up NFS and not >2GB files with a modern kernel and client? *scratching head* ;^? Thanks, Linda -- To unsubscribe, e-mail: opensuse+unsubscribe@opensuse.org For additional commands, e-mail: opensuse+help@opensuse.org
Linda Walsh wrote:
I'm exporting some files (>10GB files) via NFS from a 2.6.23.1-kernel, SuSE system, and mounting on another (same kernel).
All of my files >2GB (not exactly sure if break point is at 2 or 4GB's, all the large files I look at on the source are over 20G) give stat errors over NFS. Both kernels have nfs v4 as well as v3 compiled in.
I've heard v2 has a 2GB limit, but v3 should be fine. Have been playing with tags 'nfsvers=[3 or 4]' in the /etc/exports file, but exports doesn't recognize nfsver=4, and gives out of range error message on nfsver=3 saying allowed values are Min=1, Max=2. !? I'm not sure if it is using V3 or 4 (am trying to use 'tcp' tag as well, but not using the tcp isn't the prob, as that was the default before I started looking at this problem.
I don't have problems over CIFS from a windows box, but so far, limited benchmarks show better linux-to-linux perf over NFS (assuming I stay under 2GB files, but that's sorta limiting...).
Underlying filesystem on the server box is XFS -- been serving Win clients for years, so haven't had alot of experience using NFS, recently, since >2G files became fairly common place.
Any ideas how I could be serving up NFS and not >2GB files with a modern kernel and client? *scratching head* ;^?
which NFS deamon (nfsd) are you running? also, man 8 exportfs there might be something there that helps. Sorry I can't be more specific, because don't have the means to replicate your situation at the current time. -- To unsubscribe, e-mail: opensuse+unsubscribe@opensuse.org For additional commands, e-mail: opensuse+help@opensuse.org
On Tue, 2007-11-20 at 20:25 -0800, Linda Walsh wrote:
I'm exporting some files (>10GB files) via NFS from a 2.6.23.1-kernel, SuSE system, and mounting on another (same kernel).
All of my files >2GB (not exactly sure if break point is at 2 or 4GB's, all the large files I look at on the source are over 20G) give stat errors over NFS. Both kernels have nfs v4 as well as v3 compiled in.
I've heard v2 has a 2GB limit, but v3 should be fine. Have been playing with tags 'nfsvers=[3 or 4]' in the /etc/exports file, but exports doesn't recognize nfsver=4, and gives out of range error message on nfsver=3 saying allowed values are Min=1, Max=2. !? I'm not sure if it is using V3 or 4 (am trying to use 'tcp' tag as well, but not using the tcp isn't the prob, as that was the default before I started looking at this problem.
I don't have problems over CIFS from a windows box, but so far, limited benchmarks show better linux-to-linux perf over NFS (assuming I stay under 2GB files, but that's sorta limiting...).
Underlying filesystem on the server box is XFS -- been serving Win clients for years, so haven't had alot of experience using NFS, recently, since >2G files became fairly common place.
Any ideas how I could be serving up NFS and not >2GB files with a modern kernel and client? *scratching head* ;^?
Thanks, Linda
I routinely move files 3 to 4 Gbytes from xfs to jfs system with no problems. I am running the 2.6.22 kernel thugh and using open suse 10.3 Could you list the /etc/exports file and the rpms associated with the nfs server and the nfs client. Another thing you could do is to enable automount and uncomment the /net. There you can try to copy a large file say your machine is "hello" via "cp /net/hello/big.file /net/hello/big,file.copy" This will help to determine if your nfs server is properly configured. -- Joseph Loo jloo@acm.org -- To unsubscribe, e-mail: opensuse+unsubscribe@opensuse.org For additional commands, e-mail: opensuse+help@opensuse.org
FYI -- examining nfs versions I found I was using a user-side nfs-client. Am trying "switching" to server-side...but keep getting side-tracked on other problems. So far haven't gotten kernel-based nfs server to work remotely -- works "locally" (using autofs & /net/"server"/filesys), but on client -- if I restart "nfsserver" & try /net/server on client, client will create mount points (and showmount -e from client works). But after the mount points are created, if I try to access any, the mount-points get dynamically removed as I access each one, and showmount -e <server> returns: rpc mount export: RPC: Unable to receive; errno = Connection refused; The problem seems party on server -- since if I /etc/rc.d/nfsserver restart, showmount will start working "again", and I can get an automount from a client to remount all of the server-exported filesystems. I keep trying different options, but so far, no luck. Haven't responded, since I keep thinking I should be able to fix this -- it's just a "config" problem. Sigh... Joseph Loo wrote:
I routinely move files 3 to 4 Gbytes from xfs to jfs system with no problems. I am running the 2.6.22 kernel thugh and using open suse 10.3
Could you list the /etc/exports file and the rpms associated with the nfs server and the nfs client.
/etc/exports: /backups astara(async,rw,no_root_squash,no_subtree_check) /Share astara(async,rw,no_root_squash,no_subtree_check) /Share 192.168.3.0/255.255.255.0(async,ro,no_subtree_check) /home astara(async,rw,no_root_squash,no_subtree_check) ==== client & server running rpms: nfs-kerenel-server-1.1.0-8 nfsidmap-0.20-21 nfs-client-1.1.0-8;
Another thing you could do is to enable automount and uncomment the /net. There you can try to copy a large file say your machine is "hello" via "cp /net/hello/big.file /net/hello/big,file.copy" This will help to determine if your nfs server is properly configured.
As mentioned in passing above -- I have enabled this -- it works on the server, but not from the client -- on client, will mount the exported filesystems, but I it acts like it is the remote mountd (RPC:Unable to receive") on server that seems to be failing. I'm trying to see why mountd appears to be dying. Haven't used nfs4 enough to know "normal", but I see 4*[nfsd] but only 1 [nfsd4]. Not sure if that is right or not... -- To unsubscribe, e-mail: opensuse+unsubscribe@opensuse.org For additional commands, e-mail: opensuse+help@opensuse.org
On Mon, 2007-11-26 at 16:56 -0800, Linda Walsh wrote:
FYI -- examining nfs versions I found I was using a user-side nfs-client. Am trying "switching" to server-side...but keep getting side-tracked on other problems.
So far haven't gotten kernel-based nfs server to work remotely -- works "locally" (using autofs & /net/"server"/filesys), but on client -- if I restart "nfsserver" & try /net/server on client, client will create mount points (and showmount -e from client works). But after the mount points are created, if I try to access any, the mount-points get dynamically removed as I access each one, and showmount -e <server> returns: rpc mount export: RPC: Unable to receive; errno = Connection refused;
The problem seems party on server -- since if I /etc/rc.d/nfsserver restart, showmount will start working "again", and I can get an automount from a client to remount all of the server-exported filesystems.
I keep trying different options, but so far, no luck. Haven't responded, since I keep thinking I should be able to fix this -- it's just a "config" problem. Sigh...
Joseph Loo wrote:
I routinely move files 3 to 4 Gbytes from xfs to jfs system with no problems. I am running the 2.6.22 kernel thugh and using open suse 10.3
Could you list the /etc/exports file and the rpms associated with the nfs server and the nfs client.
/etc/exports: /backups astara(async,rw,no_root_squash,no_subtree_check) /Share astara(async,rw,no_root_squash,no_subtree_check) /Share 192.168.3.0/255.255.255.0(async,ro,no_subtree_check) /home astara(async,rw,no_root_squash,no_subtree_check) ==== client & server running rpms: nfs-kerenel-server-1.1.0-8 nfsidmap-0.20-21 nfs-client-1.1.0-8;
Another thing you could do is to enable automount and uncomment the /net. There you can try to copy a large file say your machine is "hello" via "cp /net/hello/big.file /net/hello/big,file.copy" This will help to determine if your nfs server is properly configured.
As mentioned in passing above -- I have enabled this -- it works on the server, but not from the client -- on client, will mount the exported filesystems, but I it acts like it is the remote mountd (RPC:Unable to receive") on server that seems to be failing.
I'm trying to see why mountd appears to be dying. Haven't used nfs4 enough to know "normal", but I see 4*[nfsd] but only 1 [nfsd4]. Not sure if that is right or not...
Did you enable the server on startup? You can enable the nfs server to start on run level 5. YAST->System->system services (run level). Your exports looks okay. Remember firewalls will block the nfs server being access from the client. -- Joseph Loo jloo@acm.org -- To unsubscribe, e-mail: opensuse+unsubscribe@opensuse.org For additional commands, e-mail: opensuse+help@opensuse.org
Joseph Loo wrote:
On Mon, 2007-11-26 at 16:56 -0800, Linda Walsh wrote:
I'm trying to see why mountd appears to be dying. Haven't used nfs4 enough to know "normal", but I see 4*[nfsd] but only 1 [nfsd4]. Not sure if that is right or not...
Did you enable the server on startup? You can enable the nfs server to start on run level 5. YAST->System->system services (run level).
Your exports looks okay.
Remember firewalls will block the nfs server being access from the client.
Problem is(was) in the "rpc.mountd" included in suse10.3. It was dying shortly after any client did an "ls /net/<server>" of the server with err: "rpc.mountd: symbol lookup error: rpc.mountd: undefined symbol: blkid_probe_all_new" I "backgraded" the rpm to the nfs-utils from suse10.2 and it's now working. (>2GB files visible now as well) So initial problem was I had 'some' user-space nfs-server that didn't support the >2GB files, then tried nfs-server rpm from 10.3 -- that led to an undefined symbol error. So I've installed nfs-utils from suse10.2 and its now working. Thanks for the help. Wouldn't have thought to try /net/server as a debug step. To find the error message, I finally ran rpc.mountd in foreground with debugging messages turned on. -- To unsubscribe, e-mail: opensuse+unsubscribe@opensuse.org For additional commands, e-mail: opensuse+help@opensuse.org
participants (3)
-
Aaron Kulkis -
Joseph Loo -
Linda Walsh