[lvs-users] Outdated info about SH scheduler weights

[Ferenc Wágner]

Hi,

Several Internet resources state that the weight of a realserver with
the SH scheduler is half the number of its allowed connections.  For
example:

* http://kb.linuxvirtualserver.org/wiki/Source_Hashing_Scheduling, which
  links to an
* archived mailing list thread starting at
  http://archive.linuxvirtualserver.org/html/lvs-users/2006-06/msg00001.html
* http://www.austintek.com/LVS/LVS-HOWTO/HOWTO/LVS-HOWTO.ipvsadm.html#SH-scheduler 

Looking at the code, this might have been true in Linux 2.4, but seems
to not be the case since 2.6.12 at least (where the Git history starts).
On the other hand, SH weights must be kept low to fit the 256 buckets
together since 76ad94fc (Linux 3.3), which is a difficult to handle
conflict if the weights still represented connection limits.

People with relevant knowledge, please comment on the above.
People with the necessary power, please fix the documentation.
-- 
Thanks,
Feri