NFS errors
Hello!
My computer uses another server to boot and run with NFS root. During normal work I notice strange NFS errors.
Please tell me what causes these errors and how I can resolve them.
Example 1:
I executed these commands directly after another.
cotto@carsten-otto:/home/cotto/.mozilla/firefox$ l ls: .: Permission denied cotto@carsten-otto:/home/cotto/.mozilla/firefox$ l total 8.0K drwx------ 6 cotto users 4.0K Jul 24 21:03 eir51v0r.default ?--------- ? ? ? ? ? la2v5nug.default ?--------- ? ? ? ? ? pluginreg.dat -rw-r--r-- 1 cotto users 94 Jul 24 21:03 profiles.ini ?--------- ? ? ? ? ? v1q37ws3.default cotto@carsten-otto:/home/cotto/.mozilla/firefox$ l total 20K drwx------ 6 cotto users 4.0K Jul 24 21:03 eir51v0r.default drwx------ 3 cotto users 4.0K Jul 24 21:03 la2v5nug.default -rw------- 1 cotto users 286 Jul 24 20:57 pluginreg.dat -rw-r--r-- 1 cotto users 94 Jul 24 21:03 profiles.ini drwx------ 6 cotto users 4.0K Jul 24 21:04 v1q37ws3.default
Example 2:
GAIM shows "Could not read file: somewhere.txt" when scrolling through the list of archived conversations.
Example 3:
Firefox crashes a lot during normal surfing, sometimes even the profile gets lost.
Example 4:
I get a lot of "do_vfs_lock: VFS is out of sync with lock manager!" during the day.
Example 5:
Some ./configure && make runs do not complete, because files are missing. Running the command in a chroot locally on the server (which has the disk drives) everything works.
The client (without disk drives) uses 2.6.17.6, but I have this problem with several kernel versions.
The kernel is running 2.6.18-rc1-mm2, but here again the problem occurs with several versions.
The client runs Debian testing, the server runs Debian stable (Sarge). The network connection in between is:
(Client) Intel e1000 PCI <-> HP Procurve 2824 (Gigabit) <-> Intel e1000 PCIe (Server)
x:~# rpcinfo -p server
Program Vers Proto Port
100000 2 tcp 111 portmapper
100000 2 udp 111 portmapper
100003 2 udp 2049 nfs
100003 3 udp 2049 nfs
100003 4 udp 2049 nfs
100003 2 tcp 2049 nfs
100003 3 tcp 2049 nfs
100003 4 tcp 2049 nfs
100021 1 udp 32768 nlockmgr
100021 3 udp 32768 nlockmgr
100021 4 udp 32768 nlockmgr
100021 1 tcp 43642 nlockmgr
100021 3 tcp 43642 nlockmgr
100021 4 tcp 43642 nlockmgr
100005 1 udp 779 mountd
100005 1 tcp 782 mountd
100005 2 udp 779 mountd
100005 2 tcp 782 mountd
100005 3 udp 779 mountd
100005 3 tcp 782 mountd
100024 1 udp 877 status
100024 1 tcp 880 statusI enabled nfs4 in the kernels, but do not use it yet (as a lot of changes are needed in userspace).
mount shows:
server:/export on / type nfs (rw,soft,tcp,intr)
Thanks a lot,
Signature · 3 lines
Carsten Otto [email protected] www.c-otto.de
Quoted 76 lines · Carsten Otto
Hello!
My computer uses another server to boot and run with NFS root. During normal work I notice strange NFS errors.
Please tell me what causes these errors and how I can resolve them.Example 1:
I executed these commands directly after another.cotto@carsten-otto:/home/cotto/.mozilla/firefox$ l ls: .: Permission denied cotto@carsten-otto:/home/cotto/.mozilla/firefox$ l total 8.0K drwx------ 6 cotto users 4.0K Jul 24 21:03 eir51v0r.default ?--------- ? ? ? ? ? la2v5nug.default ?--------- ? ? ? ? ? pluginreg.dat -rw-r--r-- 1 cotto users 94 Jul 24 21:03 profiles.ini ?--------- ? ? ? ? ? v1q37ws3.default cotto@carsten-otto:/home/cotto/.mozilla/firefox$ l total 20K drwx------ 6 cotto users 4.0K Jul 24 21:03 eir51v0r.default drwx------ 3 cotto users 4.0K Jul 24 21:03 la2v5nug.default(quoted diff) +0
-2-rw------- 1 cotto users 286 Jul 24 20:57 pluginreg.dat -rw-r--r-- 1 cotto users 94 Jul 24 21:03 profiles.inidrwx------ 6 cotto users 4.0K Jul 24 21:04 v1q37ws3.default
Example 2:
GAIM shows "Could not read file: somewhere.txt" when scrolling through the list of archived conversations.Example 3:
Firefox crashes a lot during normal surfing, sometimes even the profile gets lost.Example 4:
I get a lot of "do_vfs_lock: VFS is out of sync with lock manager!" during the day.Example 5:
Some ./configure && make runs do not complete, because files are missing. Running the command in a chroot locally on the server (which has the disk drives) everything works.The client (without disk drives) uses 2.6.17.6, but I have this problem with several kernel versions.
The kernel is running 2.6.18-rc1-mm2, but here again the problem occurs with several versions.
The client runs Debian testing, the server runs Debian stable (Sarge). The network connection in between is:
(Client) Intel e1000 PCI <-> HP Procurve 2824 (Gigabit) <-> Intel e1000 PCIe (Server)x:~# rpcinfo -p server Program Vers Proto Port 100000 2 tcp 111 portmapper 100000 2 udp 111 portmapper 100003 2 udp 2049 nfs 100003 3 udp 2049 nfs 100003 4 udp 2049 nfs 100003 2 tcp 2049 nfs 100003 3 tcp 2049 nfs 100003 4 tcp 2049 nfs 100021 1 udp 32768 nlockmgr 100021 3 udp 32768 nlockmgr 100021 4 udp 32768 nlockmgr 100021 1 tcp 43642 nlockmgr 100021 3 tcp 43642 nlockmgr 100021 4 tcp 43642 nlockmgr 100005 1 udp 779 mountd 100005 1 tcp 782 mountd 100005 2 udp 779 mountd 100005 2 tcp 782 mountd 100005 3 udp 779 mountd 100005 3 tcp 782 mountd 100024 1 udp 877 status 100024 1 tcp 880 status
Quoted 7 lines
I enabled nfs4 in the kernels, but do not use it yet (as a lot of changes are needed in userspace).
mount shows:
server:/export on / type nfs (rw,soft,tcp,intr)Thanks a lot,
Have you set 'no_subtree_check' in your export options on the server? If not, please try doing so.
cheers,
Trond
Hi!
Quoted 2 lines · Trond Myklebust
Have you set '
no_subtree_check' in your export options on the server? If not, please try doing so.
Oh. But Neil almost told me that :) That seems to fix a similar problem for me at least:
http://thread.gmane.org/gmane.linux.kernel/426931
Now I'm not seeing any weirdness between two 2.6.18-rc2 hosts. (No spurious "Reducing readahead size to 28/4/0K" either). I still don't get how that patch that Neil mentioned did not affect 2.6.17 (which had that), but does effect any 2.6.18-rcX.
Janos
Quoted 2 lines · Trond Myklebust
Have you set '
no_subtree_check' in your export options on the server? If not, please try doing so.
No problem so far (in the last few days), seems to work. Thanks!
Signature · 3 lines
Carsten Otto [email protected] www.c-otto.de