Oh, and it reports server version "1.2.3-yourmom" (for Whitaker).
Then in responds to queries:
bradfitz@bini:~$ mysql -hbrad -P3308 Welcome to the MySQL monitor. Commands end with ; or \g. Your MySQL connection id is 1337 to server version: 1.2.3-yourmom Type 'help;' or '\h' for help. Type '\c' to clear the buffer. mysql> select your_mom(); ERROR 1045: No access to my mom
Yup yup, workin' hard, workin' hard.
The real goal, of course, is twofold:
Goal 1) for our "global cluster" which is pretty small and all the data fits on a machine:
load balancing mysql proxy. track all machine's replication position and never direct selects to machines that aren't totally caught up. this way application thinks it's talking to a single machine, but really selects are load balanced without side-effects like out-of-date slaves. ideally then the proxies would communicate amongst themselves and elect a master, then the master would be able to do stuff like manage master election amongst the mysql machines, reparenting nodes against each other, copying their data around when a new node comes online, etc.
I think a dozen vendors do this, but what fun is that? Plus there should be some open source load-balancing mysql proxy out there.
Goal 2) for our user clusters which take a lot of disk space:
distributed database. that whole MogileFS + user per db thing, all users on 2 or 3 random machines, replication going crazy everywhere. the proxy would have to keep all their replication positions in sync.
Obviously goal 2 is a lot harder, so goal 1 is a good way to learn the MySQL protocol better and get something usable for most people (including ourselves) pretty quickly.
