On Tue, Apr 10, 2001 at 10:33:49AM -0500, Steve Lord wrote:
> We are aware of your problem, can you fill in a few more details such as
> are you running nfs v2 or v3 exports and what do you characterize as
> very heavy load, nfsstat output might be interesting after a few hours of
> uptime. Client quantities and types would be good to.
>
> I am currently working another nfs bug which only shows visible manifestation
> on solaris clients, but appears to be bad packet generation on a readdir.
I'm running NFS v3 exports to 4 Linux clients. Each client hammers on the
server by reading random 3MB files (2 processes on each client machine).
One of the clients also writes random-sized (upto 2MB) files, and periodically
deletes the old ones.
Here's the output of the following commands just a few seconds before the
oops:
% cat /proc/fs/pagebuf/stat
pagebuf 7098871 4525960 2572922 1891 43 175769 7164056 284569
% cat /proc/fs/xfs/stat
extent_alloc 28969 3916897 28637 3916006
abt 168904 588465 62088 62084
blk_map 199068226 9888984 199345 9888983 199677 209156531 167391241
bmbt 786 4667 376 399
dir 598330 199074 199041 272
trans 27996 654700 171057
ig 399256 398081 0 1175 0 0 0
log 81992 3798058 410 38929 29028
push_ail 853753 0 0 0 0 0 0 0 0 0
xstrat 28951 0
rw 25280224 89246733
attr 0 0 0 0
qm 0 0 0 0 0 0 0 0
icluster 164155 37791 93328
vnodes 1174 200215 199041 995716 116996462 199041 199041 0
xpc 16043536384 103136127363 365554618368
Next time around, I'll be running nfsstat in a loop. The problem manifests
itself after 15 hours or so, so the wait is agonizing. :-)
Thanks for the help, I didn't mean to sound pushy :-)
Ajay
|