Bug 189311

Description of problem:
NFS sunrpc module allows one to specify sunrpc.tcp_slot_table_entries 
parameter.
This parameter is set to default of 16 in sunrpc module AFAIK.
This parameter can be put in /etc/sysctl.conf. Putting a value of say 64 and 
rebooting the machine causes the value of this parameter as observed by 
sysctl -a | grep sunrpc to show up as 16.
When setting it up manually with sysctl -w sunrpc.. it works fine but if put in
sysctl.conf it doesn't.



Version-Release number of selected component (if applicable):

Redhat Enterprise Linux 4 update 3
How reproducible:
Description of the problem is sufficient to explain the reproduction


Steps to Reproduce:
1.
2.
3.
  
Actual results:


Expected results:


Additional info:

This problem was observed at highly escalated performance problem at Yahoo a 
NetApp customer that deploys their application on Redhat linux on Oracle 
database on opteron hardware.

We have observed that from the init script the sysctl is invoked only in 
S10network script. 
S18/S19 level rcp init script issues a modprobe sunrpc which probably load the 
sunrpc module. It is likely that S10network sets this value correctly but 
sunrpc module upon load up causes this value to be reset back to 16.

A simple workaround of adding /sbin/sysctl -p line in S25netfs has been 
provided to yahoo. But they would like a fix eventually from Red Hat.