============================================ #opendaylight-devforum5: release for lithium ============================================ Meeting started by colindixon at 17:05:56 UTC. The full logs are available at http://meetings.opendaylight.org/opendaylight-devforum5/2014/release_for_lithium/opendaylight-devforum5-release_for_lithium.2014-09-29-17.05.log.html . Meeting summary --------------- * current release pipeline (colindixon, 17:06:14) * mlemay talks about how we currently do releases, all projects go through milestones at the same time, hitting the same things at the same dates (colindixon, 17:07:34) * his point seems to be that inter-project dependencies mean that we really should have some delay between when the milestone for projects based on their dependency relationship (colindixon, 17:09:59) * radical idea: no more simultaneous release (colindixon, 17:10:19) * instead have gates, e.g., based on lack of bugs, code coverage, etc. (colindixon, 17:10:47) * rovarga says that currently one big problem is that we have no visibility into project inter-dependencies (colindixon, 17:11:17) * colindixon points out that this manifested as not being able to tell who meant to expose what APIs and who meant to consume what APis to track dependencies, reveal when APIs are deprecated, not yet stable, etc. (colindixon, 17:14:51) * above was partially stealing from mlemay (not all colindixon) (colindixon, 17:15:12) * goal is to fail early, catch CI failures faster for integration work (colindixon, 17:17:48) * minor aside about docs and how to do docs, with lots of ideas, probably too much to cover here (colindixon, 17:18:25) * colindixon talks about docs being managed like code (dfarrell07, 17:20:40) * discussion about integration jobs, projects not knowing about them, therefore not doing them (dfarrell07, 17:20:59) * tykeal suggests that we're not using our existing tools properly for int work (dfarrell07, 17:21:39) * Need to rethink how we are using Jenkins, not used siloed Jenkins systems (via tykeal) (dfarrell07, 17:22:19) * Silos are legicy issue, forced on tykeal before project even started (dfarrell07, 17:22:38) * colindixon points out that we may be jumpping into (dfarrell07, 17:23:36) * colindixon points out that we may be jumpping into "things are broken and we need new tools" too quickly (dfarrell07, 17:24:10) * colindixin talks about semantic buid versions "what was the last version (nightly, weekly) that passed these jobs" (dfarrell07, 17:25:12) * Question put as "should we have integration testing on snapshots or not". Currently just need your tests to pass, int tests happen after the fact. (dfarrell07, 17:26:04) * tangent on versions for a while (colindixon, 17:27:08) * Tic-toc relase mentioned: Relase "core", relase "projects", iterate every few months. Issues with what is "core" (dfarrell07, 17:27:32) * "Core" may mean base infra pieces (dfarrell07, 17:28:17) * colindixion Is working on wiki fixes to doc relase plan (dfarrell07, 17:28:42) * tangent on versions, running multiple versions at a time, discussion about this maybe being a bad idea (dfarrell07, 17:31:16) * Need to be able to take X vendor ODL controller, Y vendor app, they run together via colindixon (dfarrell07, 17:33:12) * Releases (tic-toc vs simul) may be very different for leaf projects vs other projects (dfarrell07, 17:34:56) * Discussion about offsets between deadlines, tic-toc is similar to this, again may be importantly different between types of projects (dfarrell07, 17:37:07) * rovarga talks about API freeze vs feature freeze, ordering of the two (dfarrell07, 17:38:18) * colindixon points out that we are a very new project, these issues smell like a product of getting projects spun up on how all release things work (dfarrell07, 17:40:09) * "Are there issues we think are fundi broken; which are ones that will are ghosts and will go away without hard work" colindixon (dfarrell07, 17:42:00) * We're kinda already doing tic-toc, just on 6 month cadence. MD-SAL introduced last release, used more this relase. (dfarrell07, 17:43:11) * colindixon notes that our problem in Helium on APIs is that this is the first time many APIs were used and relied on by the community (such as Yangtools and MD-SAL) apis. (phrobb, 17:43:48) * noted that this kind of API evolution will probably occur for 3 or 4 more releases (phrobb, 17:44:22) * If we have infra that is maybe-not-working, why are we buidling on top of it? Need to focus on infra projects. cdub (dfarrell07, 17:44:51) * cdub asks why we are continuing to build out features with the foundation is not strong. (phrobb, 17:45:03) * many +1s to cdub's "make foundation strong" (dfarrell07, 17:45:38) * colindixon asks "how many have lost sleep due to release?" (phrobb, 17:47:46) * "what are the problems that caused us to loose sleep" (dfarrell07, 17:48:20) * seeking agreement on what the problems are, colindixon is (dfarrell07, 17:50:36) * one issue is version, simul release, how we integrate rovarga (dfarrell07, 17:51:02) * Rob A talks about relase vehicle being < getting solid foundation features (dfarrell07, 17:51:53) * moving to Karaf was a poin point, many +1s (dfarrell07, 17:52:06) * Ryan points out that it's not just moving to Karaf, it's moving to new *anything critical* (dfarrell07, 17:52:33) * infra needs to come early in relase cycle, rovarga says (thinking about Karaf) (dfarrell07, 17:53:21) * Infra needs to be agreed on at summit, 3-4 months in advance of release rovarga (dfarrell07, 17:53:48) * "Karaf will be only relase vehicle" was only decided at least TSC call, Lori Jakab says (dfarrell07, 17:54:20) * phrobb points out that all of the above applies to docs, generally anything new, needs to be done early in cycle (dfarrell07, 17:55:20) * many +1s to suffering from docs (dfarrell07, 17:55:34) * Docs that have been vetted by others and linked from somewhare sane before we even talk about a feature being "ready", colindixon (dfarrell07, 17:56:26) * schedule is badly broken, all things that need to be in next relase signed on during last release, cdub says (dfarrell07, 17:58:17) * maybe waisting effort for new, less critical, projects to require lots of "we're going to do this" docs, Rob A (dfarrell07, 18:00:01) * cdub: end of release should be boring, should be moving to next release (dfarrell07, 18:00:19) * Incubaged projects vs stable/critical things need diff release cycles, foundation needs focus and to be a solid release (dfarrell07, 18:02:55) * we are trying to put infrastructure on a more aggressive release schedule (rovarga, 18:04:03) * such that we deliver key pieces of yangtools and mdsal early in the release cycle (rovarga, 18:04:09) * "Release in expermentail once before can even be release in non-exp mode" colindixon (dfarrell07, 18:04:12) * what we need better handle on (and hope to getout of the design summit) is the direction infrastructure _users_ need us to do (rovarga, 18:04:16) * discussion about features vs projects vs more abstract "component" (dfarrell07, 18:05:38) * and how the release discussion may be different between them (dfarrell07, 18:05:58) * out of time (dfarrell07, 18:06:05) Meeting ended at 18:06:23 UTC. People present (lines said) --------------------------- * dfarrell07 (55) * colindixon (15) * tykeal (8) * rovarga (6) * odl_meetbot (5) * phrobb (4) * regXboi (3) * ChrisPriceAB (1) * readams (1) Generated by `MeetBot`_ 0.1.4