| *** os-chef-bot has joined #openstack-chef | 02:38 | |
| *** os-chef-bot has quit IRC | 03:12 | |
| *** mrodden has quit IRC | 05:23 | |
| *** mrodden has joined #openstack-chef | 05:25 | |
| *** eroux has joined #openstack-chef | 06:42 | |
| *** eroux has quit IRC | 06:58 | |
| *** openstackgerrit has quit IRC | 08:03 | |
| *** j^2 has quit IRC | 08:52 | |
| *** j^2 has joined #openstack-chef | 08:52 | |
| *** ChanServ sets mode: +o j^2 | 08:52 | |
| jklare | sc` hi, i just looked though some parts of the oslo code and the docs an could not find much about etcd and its usage | 12:14 |
|---|---|---|
| jklare | sc` can you point me to something that shows how we will utilise etcd after adding it? | 12:14 |
| sc` | jklare: that's all tbd as far as i can tell. when i was in boston, EmilienM headed a discussion about adding etcd as an assumed base service for Future Use - i took that as, as long as it's there for use, that satisfies the requirement | 12:31 |
| sc` | https://etherpad.openstack.org/p/BOS-etcd-base-service | 12:32 |
| sc` | this was the initial proposal back earlier this year | 12:32 |
| sc` | even the devstack changes, from what i've found thus far, only seem to make it available right now | 12:33 |
| sc` | a typical corporate downstream that doesn't deploy directly from github will change the recipe to suit their own needs, as it is | 12:34 |
| jklare | sc` mhhh... not sure we want to add it right now if there is not even one project that can utilise it | 12:36 |
| jklare | sc` i mean there is not even code for it in oslo.conf | 12:36 |
| sc` | why not? i'm working on calico independently, and it will suck in etcd as well as openstack cookbooks | 12:37 |
| sc` | k8s deployers generally already have it | 12:38 |
| sc` | gotta keep up with the times to be not "dead" | 12:38 |
| sc` | https://github.com/openstack/governance/blob/master/reference/base-services.rst also says etcd is an assumed service to exist, with no definition of a particular release. being that boston was the kickoff of pike, it made sense to at least provide close to feature parity with devstack. that is how our stuff is primarily consumed before putting it to metal, as a devstack-esque chef blob of fun | 12:41 |
| sc` | i'm not just being stubborn :) | 12:42 |
| *** openstackgerrit has joined #openstack-chef | 12:43 | |
| openstackgerrit | Merged openstack/cookbook-openstack-identity master: Drop token-flush cronjob https://review.openstack.org/495839 | 12:43 |
| jklare | sc` i mean we can add it, but recipes that are not used tend to be also not tested and therefore not maintained | 12:44 |
| sc` | quite so. inspec is seen as a successor to testing such cases, and common looks like a good candidate for inspec for the queens branch | 12:45 |
| sc` | i didn't want to add a big blob of untouchable code, so i went with the basic "give me etcd" that the cookbook provides | 12:46 |
| sc` | it started that way | 12:46 |
| sc` | but cooler heads prevailed | 12:46 |
| sc` | it doesn't make a lot of sense to bake something in to use it Eventually(tm), but i'm not going to start railing against etcd. it's like railing against systemd | 12:47 |
| sc` | the time for dissenting views is about four months ago on etcd, sadly | 12:49 |
| sc` | in the room, the discussions were about not wanting to manage configs, and instead eventually push them all into etcd | 12:50 |
| sc` | i was there :) | 12:50 |
| sc` | emagana was also there, providing a better accented dissenting view than i could utter, but the theme was definitely "we don't like config files" | 12:58 |
| jklare | sc` i like the idea of etcd as to store some config, but i would like some code that could actually do that even more :) | 13:19 |
| sc` | yeah :) | 13:21 |
| sc` | it's a nebulous change, i get it | 13:21 |
| sc` | looking further into devstack, it used cinder as the canary service. perhaps i can do similar | 13:30 |
| sc` | the details are just difficult to tease out | 13:30 |
| sc` | okay, i found the magic string | 13:33 |
| sc` | https://docs.openstack.org/ocata/config-reference/block-storage/samples/cinder.conf.html has backend_url for etcd. i'll have to check the other services, but something can be shipped | 13:33 |
| sc` | telemetry and block-storage can be converted to etcd | 13:36 |
| sc` | config options exist as of ocata | 13:36 |
| sc` | jklare: would one or both be agreeable? | 13:36 |
| sc` | if i'm reading things correctly, it's a matter of adding in the right conf attributes | 13:37 |
| jklare | sc` i am not sure how this will work without some code in oslo.config, but i might have overlooked the magic part here | 14:07 |
| jklare | sc` in the end i am fine with adding etcd if we use it for at least one config option in at least one service | 14:08 |
| jklare | sc` so that we at least realise if it works or not | 14:08 |
| jklare | sc` just starting another service that is never used seems wrong imho | 14:08 |
| sc` | fair enough. it was proposed and agreed upon that the lightness of etcd was acceptable, and several of the voices in the room were operators of large fleets | 14:18 |
| sc` | i'll be working on getting cinder working, to be to at least devstack in terms of supportability | 14:20 |
| sc` | i misspoke about telemetry. needs more time under the broiler | 14:21 |
| sc` | https://pypi.python.org/pypi/etcd3gw is needed for etcd3, it seems | 14:52 |
| *** jklare has quit IRC | 21:44 | |
| *** calbers has quit IRC | 21:45 | |
| *** jklare has joined #openstack-chef | 21:51 | |
| *** calbers has joined #openstack-chef | 21:51 | |
Generated by irclog2html.py 2.15.3 by Marius Gedminas - find it at mg.pov.lt!