[16:47:25] <ToAruShiroiNeko_>	 could I ask for a uick uery
[16:47:40] <ToAruShiroiNeko_>	 actually I will ask in research
[16:59:58] <halfak>	 Hey folks.  Google hangout today
[17:02:11] <halfak>	 ToAruShiroiNeko_, ^
[23:39:59] <awight>	 halfak: Hopefully you're not still working.  But if so, I'm puzzling about the race condition between the job that refreshes model info and the cache invalidation.
[23:41:07] <awight>	 I guess we can let normal cache invalidation leave records < 1 day old alone... and only purge those if some kind of --harder flag is passed...
[23:41:34] <legoktm>	 --no-really-please-i-mean-it
[23:45:51] <awight>	 lol
[23:46:56] <awight>	 legoktm: The situation is that we don't get the model version inlined with score responses, so there's always the potential for marking scores with an outdated model version.
[23:47:00] <awight>	 oh... facepalm
[23:47:27] <awight>	 we already talked this through--the cached score records will have an *older* version than they should, it's fine to wipe them out and reload.