[Gluster-users] about ulimit -n 1M

[Wei Dong]

Hi,

I noticed that the glusterfs client tries to set ulimit -n to 1M.  When 
I run booster with non-privileged user, the following line appears 
several times in the log file:

[2009-09-14 09:15:22] W [client-protocol.c:6010:init] brick-0-0-0: 
WARNING: Failed to set 'ulimit -n 1M': Operation not permitted

When I run it with root, there's no such complaint even though 
/proc/sys/fs/file-max shows a value of 65536.  For normal user, it seems 
the default maximum of RLIMIT_NOFILE is 4096, which is a much smaller 
value than that is required by glusterfs.

I'm wondering what's the impact of RLIMIT_NOFILE has to the performance 
and scalability of glusterfs.  Does glusterfs in deed needs to open a 
large number of files when my DHT aggregates ~100 machines and ~500 
volumes?  I'm not sure if I should increase RLIMIT_NOFILE to a large 
number for every user when everyone needs to run booster in order to 
access glusterfs.

Thanks,

- Wei