spymemcached, ketama, and adding new servers

[Dave Peckham]

Same (bad) results. I have uploaded the test project to
http://s3.amazonaws.com/00WYH5EB2EB0ECC08PG2-spyketamatest/spyketamatest.tgz

First I start my memcache instances (test is somewhat hard-coded to
these ports, but easily modifiable):

> for port in `jot 10 11211`; do memcached -p $port -d; done

Then I run the tests. Here is the output I get:

> mvn clean test
(sorry that the spy code isn't in any maven repo that I could find)

Testing with 100000 cache numEntries.
Starting first client with the partial server list.
************* server cur_items **************
localhost/127.0.0.1:11211 --> 11146
localhost/127.0.0.1:11212 --> 10983
localhost/127.0.0.1:11213 --> 11128
localhost/127.0.0.1:11214 --> 11214
localhost/127.0.0.1:11215 --> 10959
localhost/127.0.0.1:11216 --> 11230
localhost/127.0.0.1:11217 --> 11125
localhost/127.0.0.1:11218 --> 11162
localhost/127.0.0.1:11219 --> 11057
******************** end ********************
Got 99999 hits on initial server list.
Shutting down client...
Starting another client with the full server list.
************* server cur_items **************
localhost/127.0.0.1:11211 --> 11146
localhost/127.0.0.1:11212 --> 10983
localhost/127.0.0.1:11213 --> 11128
localhost/127.0.0.1:11214 --> 11214
localhost/127.0.0.1:11215 --> 10959
localhost/127.0.0.1:11216 --> 11230
localhost/127.0.0.1:11217 --> 11125
localhost/127.0.0.1:11218 --> 11162
localhost/127.0.0.1:11219 --> 11057
localhost/127.0.0.1:11220 --> 0
******************** end ********************
Got 10016 hits after adding the nth server.
Cache disruption % = 0.8998389983899839

I must be doing something wrong, but I can't see it. If need be, I'll
tear into the spy code, but I'm hoping that Dustin might be able to
shed some light before I undertake that. I don't mind doing it though.

Thanks all,
Dave