19:20:47 <kgriffs> #startmeeting Marconi
19:20:48 <openstack> Meeting started Thu May  2 19:20:47 2013 UTC.  The chair is kgriffs. Information about MeetBot at http://wiki.debian.org/MeetBot.
19:20:49 <openstack> Useful Commands: #action #agreed #help #info #idea #link #topic #startvote.
19:20:51 <openstack> The meeting name has been set to 'marconi'
19:21:04 <kgriffs> https://wiki.openstack.org/wiki/Meetings/Marconi#Agenda
19:21:33 <kgriffs> first off, any last-minute additions for the agenda today?
19:22:27 <oz_akan> hi
19:22:50 <kgriffs> hey Oz, glad you could make it
19:23:09 <kgriffs> today we'll be discussing several blueprints to get everyone's feedback
19:23:23 <kgriffs> then we'll have an open discussion (time permitting)
19:23:31 <kgriffs> #topic State of the project
19:24:30 <kgriffs> So, first off, we've been hard at work getting our blueprints in order, plowing through our patch backlog (gerrit is looking much cleaner now), and aligning remaining v1 work with Havana milestones.
19:24:49 <kgriffs> Trello is now set up to track those milestones, as well as the blueprints.
19:26:13 <kgriffs> If anyone would like to contribute to the project, just let me know and I'll add you to the Trello board. Right now it is a manual process; eventually we hope to migrate away from Trello, but are waiting to see if the blueprint UX can be improved.
19:26:45 <kgriffs> #info https://trello.com/b/7NLODgbr
19:28:09 <kgriffs> so, for H1 we are planning to clean up some cruft in the code, finish up the feature set for "version 1.0"
19:29:09 <kgriffs> Also, oz_akan and malini have started work on setting up a QA/reference cluster, and we will have system and load tests coming online shortly thereafter.
19:29:21 <kgriffs> oz_akan, malini: any questions, thoughts, etc. on that?
19:29:36 <malini> nope, I am good
19:29:50 <oz_akan> nothing at the moment, work in progress
19:30:02 <malini> we'll have the scripts for system & load tests added to marconi repo soon
19:30:12 <oz_akan> I am trying to automate creation of QA environment
19:30:37 <kgriffs> OK, cool. Any questions for Malini and Oz before we move on?
19:31:15 <malini> we are using Tsung for load tests..Any expertise on tht wud be welcome :)
19:31:25 <malini> non-expertise too ;)
19:33:08 <kgriffs> OK, so next, just wanted to give a shout-out to Victoria, a former OpenStack intern, who is taking over maintenance of python-marconicore.
19:34:27 <kgriffs> besides Jamie (jdprax), a couple other early contributors are moving on to other projects, but we are picking up new contributors and hope to pick up a bunch more in the near future (hint, hint)
19:35:20 * bryansd is still here in spirit!
19:35:44 <kgriffs> you guys are always free to contribute part-time, of course. :D
19:36:27 <kgriffs> so, thanks guys for all the awesome work you did to make Marconi a reality in just a couple months.
19:36:57 <kgriffs> The demo in Portland went really well, and was thanks to the hard work of a lot of folks.
19:37:15 <kgriffs> next topic...
19:37:22 <kgriffs> blueprints
19:37:50 <kgriffs> #topic Blueprint: FIFO, Message Pagination
19:38:15 <kgriffs> #info https://wiki.openstack.org/wiki/Marconi/specs/storage/message-pagination
19:38:43 <kgriffs> so, this has been a hot topic, both at the summit and over the past few weeks
19:38:57 <kgriffs> I wanted to get everyone's thoughts on the latest proposal
19:39:36 <kgriffs> basically, we want to guarantee FIFO, but only in the case of a single producer posting messages in a single queue.
19:40:09 <kgriffs> Does anyone see a practical use case for doing FIFO for multiple producers?
19:40:36 <kgriffs> What I think people are really after when they talk about wanting FIFO is a workflow service.
19:41:17 <kgriffs> While Marconi may be used *by* a workflow service, I don't think it should try to *be* a workflow service.
19:41:26 <kgriffs> Thoughts?
19:41:49 <kgriffs> (see also: https://etherpad.openstack.org/marconi-fifo)
19:46:10 <kgriffs> lively crowd today. :D
19:46:47 <malini> :d
19:47:31 <kgriffs> OK, so I guess if nobody has any immediate objections, we will run with this. Send questions/comments/concerns to the list or bring them up in #openstack-marconi
19:47:48 <kgriffs> #topic Blueprint: Grace Period for Expiration of Claimed Messages
19:48:20 <kgriffs> #info https://blueprints.launchpad.net/marconi/+spec/claimed-message-grace
19:49:08 <kgriffs> So, the idea here, is that if a consumer crashes and it's claim expires, we need the claimed messages to live a little longer beyond the claim so that another consumer has a chance to reclaim them.
19:49:50 <kgriffs> we already extend message lifetimes (if needed) so they never expire before the claim does
19:50:10 <kgriffs> the proposal is to add a grace period (seconds) to that extension.
19:50:41 <kgriffs> any objections?
19:51:07 <kgriffs> and, should this grace setting be per-queue or per-message?
19:54:44 <oz_akan> is per-message way more difficult than per-queue?
19:55:01 <oz_akan> ...to implement
19:56:16 <kgriffs> no, it's purely a matter of what would make the most sense for the user
19:57:13 <kgriffs> there may be a small optimization we could do if each message post included the grace.
19:57:37 <kgriffs> basically, the storage driver would not have to look up the queue's metadata each time to get the grace seconds
19:58:39 <kgriffs> probably not a big gain in performance (we'd have to benchmark it)
19:59:36 <kgriffs> anyone have strong feelings on whether grace should be a per-message vs. per-queue setting?
19:59:45 <kgriffs> BTW, currently TTL is per-message
20:00:41 <markwash> glance time?
20:00:46 <kgriffs> yep
20:00:50 <kgriffs> let me just wrap up Marconi
20:01:07 <markwash> take your time :-)
20:02:09 <kgriffs> OK, so it sounds like everyone is cool with the current direction of the project and the FIFO proposal. Please help spread the word and encourage your peeps to contribute. :D
20:02:37 <kgriffs> Any last items?
20:03:28 <kgriffs> P.S. - if there are any of the blueprints on the agenda that you had questions/concerns/comments about, that we didn't get to, just pop into #openstack-marconi any time to discuss.
20:04:31 <kgriffs> We have an aggressive development schedule, so we will likely move on these blueprints before the next community meeting.
20:04:38 <kgriffs> cheers,
20:04:46 <kgriffs> #endmeeting