[14:15:23] o/ dsaez and miriam [14:15:30] Anything for the Research Group agenda? [14:22:53] hey halfak, nothing on my side for today :) [14:24:54] Gotcha. I think I'll be canceling it and sending out a notice. No agenda == no meeting :) [14:31:41] :) ok [14:44:08] o/ people/ [14:44:34] o/ leila :) Good morning [14:44:45] halfak: thanks for staying on top of this meeting. [14:44:53] good morning, halfak. [14:45:35] dsaez: hola. let's talk in an hour about the evaluation of your part, ok? [14:47:10] dsaez: I do think independent of the fusion approach and whether we want to do it or not, the section alignment approach you have developed needs evaluation. [14:47:11] dsaez: re evaluation the recommendations from your part only, sure, that part is not very clear and we may or may not want to do it. We should discuss this latter part. [16:59:30] sorry - stupid beginner question, how do I know if I am in the wmf group in LDAP? [20:06:22] Research documentation time! [20:06:33] I'm here working on documenting JADE, ORES new auditing system [20:55:14] I have a puppet patch for review (https://gerrit.wikimedia.org/r/#/c/383916/) that makes it simple to configure labs instances for stats computing, forecasting, Bayesian inference via MCMC, and machine learning in Python and R. (Deep learning is a TODO.) Currently the roles are organized under "Discovery" but if halfak, leila, HaeB, dsaez, miriam, et al. would be interested in having these available, I can re-organize it so [20:55:14] they're general. [20:55:50] what do the roles do? [20:56:04] Install a basic set of libraries so a labs VM looks like a stat machine? [20:56:09] Why not use the stat machine role? [20:56:19] ottomata, ^ [20:59:26] there's stuff that comes with stat machine role (this one I think: https://github.com/wikimedia/puppet/blob/production/modules/statistics/manifests/compute.pp) that creates restrictions [21:00:51] but yeah, the roles install different sets of libraries. and it's meant for, like, spinning up an instance on wm cloud, applying a role, doing the task you want, and then deleting the instance to clear resources [21:03:49] plus these roles can install libraries from github and other places, which the stat machine role can't for security reasons [21:05:49] the patch as is it addresses my team's needs, but I just don't know if any other teams have similar needs so I'm double checking with yinz :) [21:10:08] Gotcha. Sounds like a common enough use-case. Not sure how ops is going to feel about having a puppet role that installs things from github. Is that really necessary to have in the role or can it be ad-hoc when the VM is online? [21:11:12] halfak: oh the github stuff is already in place from earlier work for our dashboards :) [21:12:03] Gotcha. So if you're spinning up a VM that won't host dashboards, you just end up installing a bunch of extra stuff? [21:13:41] halfak: what do you mean? [21:14:03] Oh I'm just imagining that you have a single role that is a catch-all that you keep re-using. [21:14:14] Right now, the stat role is sort of a catch-all for lots of different types of projects. [21:19:39] halfak: which role(s) do you use for labs VMs? [21:20:05] Mostly ores::base unless we're hosting something. [21:20:13] e.g. ores::web or ores::celery [21:20:45] Sorry ores::celery --> ores::worker [21:20:46] https://github.com/wikimedia/puppet/tree/production/modules/ores/manifests [21:20:53] ores::redis [21:20:54] etc. [21:21:46] bearloga, I need to run. Thanks for calling my attention to the role. Don't block on me. I'll keep your work in mind and maybe bring it back up to you later :) [21:21:54] o/ [21:21:59] cheers o/ [22:02:41] * leila reads [22:05:14] bearloga: I just caught up on this thread. I'm where Aaron is. Thanks for checking. :)