Low default for tcp_slot_table_entries

[Michael Gutteridge]

The value for /proc/sys/sunrpc/tcp_slot_table_entries seems unusually low on 12.04.1LTS. The default seems to be 2:

root@g:~# sysctl sunrpc.tcp_slot_table_entries
sunrpc.tcp_slot_table_entries = 2
root@g:~# sysctl sunrpc.udp_slot_table_entries
sunrpc.udp_slot_table_entries = 16

NFS documentation indicates that the default should be 16. Early indications from my own testing is that increasing this shows an improvement in performance.

[EclipseAgent (ben-kevan)]

This should be closed out. NFS clients moved to dynamic setting of the tcp_slow_table_entries as evident of a default value of 2.

[Michael Gutteridge (mrg-d)]

I concur- though I haven't yet found anything but second-hand reports that this is the case (i.e. table entries are dynamic and with the upper bound controlled by sunrpc.tcp_max_slot_table_entries).

I'm uncertain of the best way to close the issue... set state to "invalid"?