| *** jamesmcarthur has quit IRC | 00:22 | |
| *** jamesmcarthur has joined #openstack-tc | 00:23 | |
| *** jamesmcarthur has quit IRC | 00:27 | |
| *** lbragstad has quit IRC | 00:34 | |
| *** lbragstad has joined #openstack-tc | 01:01 | |
| mnaser | tc office hours :-) | 01:01 |
|---|---|---|
| lbragstad | o/ | 01:02 |
| mnaser | pretty happy of the turnout for the upcoming election honestly | 01:14 |
| fungi | office hours indeed | 01:14 |
| fungi | and yes, i'm pleased with the list of candidates as well | 01:15 |
| mnaser | i'm happy to see some of the names that came and hope this is finally when we get some apac represetnation | 01:17 |
| fungi | or at least get more apac representation | 01:20 |
| lbragstad | ++ | 01:27 |
| flwang | lbragstad: waves from NZ | 01:43 |
| lbragstad | flwang o/ glad to see you're running for TC :) | 01:44 |
| zaneb | mnaser: +1 | 01:47 |
| flwang | lbragstad: yep, i'm brave again to run for it | 01:48 |
| *** whoami-rajat has joined #openstack-tc | 02:32 | |
| *** jamesmcarthur has joined #openstack-tc | 02:37 | |
| *** ricolin has joined #openstack-tc | 03:01 | |
| *** jamesmcarthur has quit IRC | 03:21 | |
| *** jamesmcarthur has joined #openstack-tc | 03:22 | |
| *** jamesmcarthur has quit IRC | 03:27 | |
| *** jamesmcarthur has joined #openstack-tc | 03:44 | |
| *** jamesmcarthur has quit IRC | 03:59 | |
| *** jamesmcarthur has joined #openstack-tc | 04:00 | |
| *** jamesmcarthur has quit IRC | 04:05 | |
| *** jamesmcarthur has joined #openstack-tc | 04:09 | |
| *** ricolin has quit IRC | 04:18 | |
| *** jamesmcarthur has quit IRC | 04:22 | |
| *** jamesmcarthur has joined #openstack-tc | 04:25 | |
| *** ricolin has joined #openstack-tc | 04:30 | |
| *** jamesmcarthur has quit IRC | 04:30 | |
| *** lbragstad_ has joined #openstack-tc | 04:36 | |
| *** lbragstad has quit IRC | 04:37 | |
| *** jamesmcarthur has joined #openstack-tc | 04:53 | |
| *** diablo_rojo has quit IRC | 04:55 | |
| *** jamesmcarthur has quit IRC | 04:59 | |
| *** lbragstad has joined #openstack-tc | 05:23 | |
| *** lbragstad_ has quit IRC | 05:25 | |
| *** lbragstad_ has joined #openstack-tc | 05:29 | |
| *** lbragstad has quit IRC | 05:31 | |
| *** lbragstad has joined #openstack-tc | 05:41 | |
| *** lbragstad_ has quit IRC | 05:43 | |
| *** lbragstad_ has joined #openstack-tc | 05:57 | |
| *** lbragstad has quit IRC | 05:58 | |
| *** lbragstad has joined #openstack-tc | 06:20 | |
| *** lbragstad_ has quit IRC | 06:22 | |
| *** Luzi has joined #openstack-tc | 06:52 | |
| *** dangtrinhnt has quit IRC | 07:01 | |
| *** jamesmcarthur has joined #openstack-tc | 07:18 | |
| *** jamesmcarthur has quit IRC | 07:22 | |
| *** e0ne has joined #openstack-tc | 07:25 | |
| *** dangtrinhnt has joined #openstack-tc | 07:26 | |
| *** lbragstad has quit IRC | 08:15 | |
| *** jpich has joined #openstack-tc | 08:24 | |
| *** dtantsur|afk is now known as dtantsur | 08:34 | |
| *** tosky has joined #openstack-tc | 08:43 | |
| *** ricolin has quit IRC | 08:55 | |
| *** e0ne has quit IRC | 09:33 | |
| *** zaneb has quit IRC | 09:53 | |
| *** ianychoi has quit IRC | 10:59 | |
| *** e0ne has joined #openstack-tc | 11:54 | |
| *** cdent has joined #openstack-tc | 12:04 | |
| *** jamesmcarthur has joined #openstack-tc | 13:18 | |
| *** jamesmcarthur has quit IRC | 13:22 | |
| *** cdent_ has joined #openstack-tc | 13:22 | |
| *** jamesmcarthur has joined #openstack-tc | 13:22 | |
| *** ianychoi has joined #openstack-tc | 13:22 | |
| *** cdent_ has quit IRC | 13:23 | |
| *** mriedem has joined #openstack-tc | 13:28 | |
| *** ijolliffe has joined #openstack-tc | 13:29 | |
| *** jamesmcarthur has quit IRC | 13:33 | |
| openstackgerrit | Merged openstack/governance master: Create separate Placement project https://review.openstack.org/636416 | 13:40 |
| *** spsurya has joined #openstack-tc | 13:44 | |
| *** jamesmcarthur has joined #openstack-tc | 13:51 | |
| *** jamesmcarthur has quit IRC | 13:51 | |
| *** jamesmcarthur has joined #openstack-tc | 13:52 | |
| *** lbragstad has joined #openstack-tc | 14:32 | |
| *** lbragstad_ has joined #openstack-tc | 14:40 | |
| *** lbragstad has quit IRC | 14:41 | |
| *** jamesmcarthur has quit IRC | 14:53 | |
| *** lbragstad has joined #openstack-tc | 14:53 | |
| *** lbragstad_ has quit IRC | 14:54 | |
| *** lbragstad_ has joined #openstack-tc | 14:56 | |
| *** lbragstad has quit IRC | 14:58 | |
| *** lbragstad_ has quit IRC | 14:58 | |
| *** lbragstad has joined #openstack-tc | 15:01 | |
| *** Luzi has quit IRC | 15:11 | |
| *** lbragstad_ has joined #openstack-tc | 15:12 | |
| *** lbragstad has quit IRC | 15:13 | |
| *** lbragstad_ is now known as lbragstad | 15:18 | |
| *** e0ne has quit IRC | 16:01 | |
| *** e0ne has joined #openstack-tc | 16:02 | |
| *** e0ne has quit IRC | 17:04 | |
| *** dtantsur is now known as dtantsur|afk | 17:14 | |
| *** jpich has quit IRC | 17:25 | |
| *** jaosorior has quit IRC | 17:49 | |
| *** jaosorior has joined #openstack-tc | 17:49 | |
| *** diablo_rojo has joined #openstack-tc | 18:01 | |
| *** cdent has quit IRC | 18:09 | |
| dhellmann | tonyb , diablo_rojo , fungi : I have tagged the governance repo at c452ccff321f5a776e9bb0ca1540ff551bc93d7d with feb-2019-elections as requested. Please let me know if everything looks right. | 18:10 |
| fungi | dhellmann: lgtm! | 18:33 |
| fungi | thanks a ton | 18:33 |
| *** cdent has joined #openstack-tc | 18:42 | |
| *** e0ne has joined #openstack-tc | 18:50 | |
| diablo_rojo | thanks dhellmann! | 19:00 |
| dhellmann | thanks for confirming | 19:01 |
| *** e0ne has quit IRC | 19:10 | |
| *** jaosorior has quit IRC | 19:18 | |
| *** jaosorior has joined #openstack-tc | 19:28 | |
| *** e0ne has joined #openstack-tc | 19:30 | |
| mnaser | one day i'll convince some people to write an oslo messaging driver that uses a combination of etcd+http to do rpc | 19:34 |
| jroll | ++ | 19:34 |
| mnaser | and the best part is it wouldn't even be that complicated because as long as we can get an implementation inside oslo messaging going, it'd be so easy afterwards | 19:35 |
| cdent | I thought someone had already started that? | 19:35 |
| cdent | I'm not sure how far they got though | 19:36 |
| smcginnis | Tooz and etcd3gateway? | 19:36 |
| jroll | I'm just curious how well the interface maps to that model | 19:37 |
| jroll | one day I'll have the time to turn off everything else and dive in | 19:37 |
| cdent | rotflmaocoptr | 19:38 |
| cdent | "one day" | 19:38 |
| dhellmann | the best idea no one ever implemented... | 19:39 |
| mnaser | i guess i have fun weekend plans ahead of me | 19:40 |
| cdent | Do you have a vice or favorite drink/food/whatever mnaser ? | 19:40 |
| cdent | I'd like to promise you such a thing if you manage to pull it off, but want to get in the right ballpark. | 19:41 |
| mnaser | i'd probably just be happy enough if people pick it up and it becomes a standard, then i never have to run rabbitmq ever again | 19:41 |
| smcginnis | mnaser: https://github.com/dims/etcd3-gateway | 19:41 |
| *** e0ne has quit IRC | 19:41 | |
| mnaser | ou thats useful | 19:42 |
| smcginnis | In use for distributed locking, but seems like it shouldn't be too difficult to use for messaging too. | 19:43 |
| mnaser | smcginnis: the idea is that we wouldnt use it for messaging, but instead simply as a key-value store to register services and endpoints | 19:43 |
| mnaser | and they talk directly to each other instead of through $some_tool | 19:43 |
| cdent | it's probabl possible to fake rpc over etcd with watchers | 19:44 |
| smcginnis | Tricky part would be handling service restarts and such. | 19:44 |
| cdent | which would more closely align with the oslo.messaging interface | 19:44 |
| jroll | mnaser: not etcd-based, but this may be relevant to your interests as well: https://review.openstack.org/#/c/633052/ | 19:44 |
| cdent | but it might be that trying to hew to om interface is an impedance mismatch | 19:45 |
| smcginnis | jroll: Any noticable performance improvements with that? | 19:45 |
| mnaser | oh thats interesting | 19:45 |
| mnaser | json-rpc hmm | 19:45 |
| jroll | smcginnis: the patch was originally written to make a tiny standalone ironic deployment easier - I haven't had time to see how it does at scale | 19:45 |
| mnaser | wow thats really interesting | 19:46 |
| smcginnis | Hah, funny. I remember talking to jgriffith about doing something similar for a smaller cinder standalone option. | 19:46 |
| jroll | I'd also prefer a real API over json-rpc, but this got the job done | 19:46 |
| mnaser | i was pretty much thinking of pushing that a step lower down to oslo.messaging and we'd all get that benefit i guess | 19:46 |
| dhellmann | mnaser : you should look at the oslo.messaging cast() feature. it doesn't really work with point-to-point semantics, it's more of a work queue. | 19:46 |
| mnaser | dhellmann: yeah, that's an interesting thing you're bringing up here as well | 19:47 |
| dhellmann | if we were just doing point-to-point rpc someone would have removed rabbit a long time ago | 19:47 |
| mnaser | i guess the "alternative" is that we can do really do point-to-point technically | 19:48 |
| dhellmann | my idea was to use etcd as the broker, instead of rabbit. post messages to a specific place based on the topic, etc. then have clients listening there. | 19:48 |
| mnaser | we can get a list of all the services identified by name (i.e. nova-scheduler) | 19:48 |
| mnaser | and then pick one at random to send a request to | 19:48 |
| dhellmann | the zmq advocates tried the approach of using zmq to build their own broker, but that just meant you had to run *that* | 19:49 |
| dhellmann | I don't know if that's going to have the same load balancing characteristics of what we do now, but it's interesting | 19:49 |
| mnaser | technically, that's about the same thing as casting it, because usually we assume only a single consumer | 19:49 |
| dhellmann | oh, cast() can also have multiple responses, I think? maybe talk to kgiusti | 19:49 |
| mnaser | i can imagine other scenarios like periodic updates from agents to update other agents on their state | 19:49 |
| mnaser | i dont know the super internals of oslo messaging yet but im assuming there's two types of casts, one that is picked up by all agents (i.e. periodic update) and one that is picked up by one only to do work (i.e. nova-conductor asking schedulers to pick a host) | 19:50 |
| *** e0ne has joined #openstack-tc | 19:51 | |
| mnaser | it would probably be a good exercise to also look at how other tools build their architecture (say, k8s.. for messaging) | 19:56 |
| tonyb | dhellmann: Thanks. #shipit Oh wait ... :D | 19:57 |
| smcginnis | Also keep in mind active/passibe and active/active HA deployments. That can get tricky real quick if you start talking point to point communications. | 19:59 |
| dhellmann | mnaser : there are 4 methods for sending in the driver, for direct, topic, fanout, and notify. I don't remember off the top of my head how they're used by the higher-level code, though | 20:00 |
| dhellmann | smcginnis : good point. there's a reason that message brokers exist. | 20:00 |
| mnaser | dhellmann: interesting | 20:01 |
| mnaser | of course, i'm not saying that we should start building our own message broker .. but, they are a big pain point for deployers | 20:01 |
| dhellmann | hence my suggestion to put the message in etcd, which is something they're used to running with fewer complaints, and build the driver to use etcd as a broker | 20:02 |
| dhellmann | if it was easy, we would have done it a long time ago | 20:02 |
| *** jamesmcarthur has joined #openstack-tc | 20:02 | |
| cdent | Another thing to keep in mind with the way we do RPC, now, is that we tend to send around big objects, rather than signalling that a thing (kept centrally) needs something to happen | 20:04 |
| mnaser | cdent: yeah thats' the thing that's on my mind when im thinking about this | 20:04 |
| mnaser | if say, etcd is designed to move huge objects around | 20:04 |
| cdent | If we were a bit more pub/subby we could say "identifier X needs someone to take an action on it" | 20:05 |
| cdent | rather than "someone, I provide you with X, please do this thing" | 20:05 |
| mnaser | cdent: i think part of it is the fact that conductors are a thing | 20:06 |
| mnaser | so agents need to still get all the info they need because they cant talk to the db | 20:06 |
| mugsie | from what I remember k8s has issues using etcd for a queue - for larger (100+ nodes) we used to have to try and separate the events etcd from the state store one. | 20:39 |
| *** spsurya has quit IRC | 20:42 | |
| *** jamesmcarthur has quit IRC | 21:05 | |
| *** jamesmcarthur has joined #openstack-tc | 21:17 | |
| *** jamesmcarthur has quit IRC | 21:55 | |
| *** jamesmcarthur has joined #openstack-tc | 21:59 | |
| *** mriedem has quit IRC | 22:25 | |
| *** whoami-rajat has quit IRC | 22:31 | |
| *** e0ne has quit IRC | 22:32 | |
| *** gagehugo has quit IRC | 22:41 | |
| *** ijolliffe has quit IRC | 22:41 | |
| *** cdent has quit IRC | 22:44 | |
| *** jamesmcarthur has quit IRC | 23:55 | |
| *** jamesmcarthur has joined #openstack-tc | 23:56 | |
| *** mriedem has joined #openstack-tc | 23:57 | |
Generated by irclog2html.py 2.15.3 by Marius Gedminas - find it at mg.pov.lt!