Earlier  
Later
New Search
Posted Nick Remark
#openstack-nova - 2017-05-05
19:22:46 mriedem so go nuts
19:22:50 mriedem like i said, i don't get there until noon
19:23:23 mriedem looking at your flow page, we shouldn't be creating a port twice,
19:23:39 mriedem but we do POST a port and then PUT (update it with the host binding) later
19:24:14 dansmith sdague: is that rst?
19:24:24 sdague dansmith: I have rst
19:24:42 dansmith can I have rst?
19:24:43 sdague https://dague.net//testing/flow.rst
19:24:57 sdague I can also stick this in a throw away git repo if you want
19:24:59 mriedem sdague: i only see "POST /v2.0/ports.json" once
19:25:10 sdague mriedem: yeh, my bad, I glossed over
19:25:10 dansmith thanks, I'll try to trim it
19:25:21 openstackgerrit Eric Fried proposed openstack/nova master: nova.utils.get_service_url() https://review.openstack.org/458257
19:25:28 sdague dansmith: how about I do an import into a git repo
19:25:50 dansmith sdague: that seems entirely overkill to me, but as you wish
19:25:52 mriedem i'm not sure what this is: May 05 14:12:54 os4 neutron-server[24158]: DEBUG neutron.api.v2.base [req-8b4ccea7-1462-43cc-9a21-f0013b797999 demo demo] Request body: {u'port': {u'network_id': u'8356212e
19:26:58 mriedem "neutron-server starts building that network:" isn't actually building a network
19:27:03 mriedem it's finding the available network for the project
19:27:06 mriedem on which to create the port
19:27:21 mriedem we'd only create a network if we use the auto-allocated-topology (get me a network) stuff
19:27:35 dansmith sounds like mriedem wants to trim it
19:27:43 mriedem i don't,
19:27:45 mriedem i just want to nit pick it
19:28:10 mriedem i'll go back to reading the thread about stable branch LTS and n-5 upgrade support
19:28:25 sdague dansmith: what's your github id again?
19:28:29 dansmith kk7ds
19:28:52 sdague https://github.com/sdague/nova-boot-flow
19:28:55 sdague you have commit
19:29:03 sdague mriedem: if you want so as well
19:29:40 sdague dansmith: I was trying to keep all the logs around, just build the summaries when appropriate, so if you go to trim, how about just getting rid of some summaries
19:33:10 dansmith sdague: yeah, I meant trim the summaries, not the log lines
19:33:12 dansmith trim and edit the summaries I mean
19:35:50 sdague dansmith: have at, you should have full write on that repo
19:36:43 dansmith sdague: okay, I'm still scrambling with other stuff but I'll try
19:36:48 dansmith I guess that would be easy to do on the plane
19:36:56 sdague dansmith: yep, being git and all
19:41:01 cdent mriedem: maybe you should throw some downgrade support hand grenades into that thread
19:44:51 mriedem jesus
19:44:58 mriedem the features vs stability thing drives me nuts
19:45:13 mriedem no one mentioned how we talked about non-feature bug fix only releases once
19:45:43 mriedem 'openstack doesn't / can't scale', ok we'll work on that, 'but you're working on it makes the transition hard and complex and that pisses me off'
19:46:18 mriedem it's just easier to not say anything at all in these threads because they pop up every 3-6 months and die on the vine anyway
19:46:53 mriedem i did see cells v2 and placement get run through the mud a few times though, that's fun
19:51:17 mriedem i guess on one replied to the request for input on the onboarding session in the ML
19:51:17 sdague yeh, my entire point on that thread at this point is going to be "show me a gratuitous feature"
19:51:18 mriedem oh well
19:51:34 mriedem sdague: cells v2 and placement were called out as causing pain with no benefit
19:51:47 sdague I guess I missed where placement was called out on that
19:51:48 mriedem not directly replying ot you
19:51:59 sdague I did just rejump in on the cellsv2 one
19:52:45 sdague so, I will agree that there were some more bumps than would have been nice there, but compared to cells v1, which people kept deploying and getting unfixable behavior... it was a much better path
19:54:40 mriedem cdent: an anecdote really http://lists.openstack.org/pipermail/openstack-dev/2017-May/116440.html
19:54:49 mriedem "nova placement API caused a bit of a headache for deployment tools..."
19:55:00 cdent mriedem: I think you mean sdague ?
19:55:07 mriedem oh yeah
19:55:07 mriedem sorry
19:55:11 cdent s'alright
19:55:51 cdent "but consider that "leadership" is always limiting for innovation."
19:57:56 cdent I'm hopeful that the "everyone deploy with wsgi" will finally get people understanding that deploying with wsgi means "you can do whatever you want, easily, but you have to choose" instead of "that's different"
19:58:23 sdague mriedem: ok, that's a pretty soft hit honestly
19:58:41 mriedem sdague: yeah i said it was an anecdote on re-read
19:58:42 sdague and, I would agree with it, the docs lagged
19:58:51 mriedem which docs/
19:58:51 mriedem ?
19:58:58 mriedem we had pretty solid in-tree docs by rc1
19:59:04 mriedem the install guide was late
19:59:08 mriedem for both placement and cells v2
19:59:09 sdague the install guide
19:59:24 sdague because that's the bit that people on the deploy side do need
19:59:42 sdague and, I think that's fine critique of "yep, we need to do that better next time"
19:59:44 mriedem sure, but when is the last time someone from nova was writing the install guide?
20:00:06 sdague mriedem: yep, sure, a bunch of things changed since the last major service changes like this showed up
20:00:22 mriedem isn't this like the thing in newton when matt kasawara wanted the install guide in nova so we'd own it, and we said no
20:00:44 sdague mriedem: no, I think this was some collective falling through the cracks
20:01:29 sdague where there was a lot of assumptions about who all owned things, that seemed to fall appart, because the last time a new service was rolled out was conductor, and the landscape was a lot different
20:01:39 sdague and ... we kind of happened to have 2 hit at the same time
20:01:49 sdague so, stuff happens
20:02:13 mriedem i agree that not having a rushed clusterfck of a release in ocata would have been better :)
20:02:21 mriedem and i think we retrospected on that
20:02:21 sdague mriedem: right, also that
20:02:36 sdague anyway
20:02:48 mriedem but i will say i'm happy with the focus put on the in-tree docs for placement and cells v2 that a bunch of people worked on
20:03:04 mriedem in years past our bar for documentation of features in tree was non-existent
20:03:11 mriedem just like our bar for requiring testing
20:03:25 sdague my point is, there is totally legit criticism on some things that were less than ideal for service / major feature rollout.
20:03:45 sdague but that's very different than saying those are gratuitous features
20:03:47 mriedem sure, i don't disagree, and i think we/i have been as critical
20:05:49 sdague I'm just sick of "features trump stability/usability" where features are code for "stuff I don't like or understand" and "stability/usability" is code for "my personal priority". And don't want to let that one slide this time.
20:12:32 dansmith I especially like "rearchitect the entire project ... in a branch"
20:17:52 sdague yeh... I am sure the upgrade process would be very smooth there
20:28:27 cdent Is "solid governance model" code for "not letting the devs decide"?
20:34:57 sdague probably
20:35:19 sdague or, more likely "put my distguished architect in charge"
20:35:57 sdague ok... I'm saving to draft a response on that last oracle one. Because that's going to spiral out of control if I don't
20:36:04 sdague I should make sure mordred sees it
20:36:38 sdague because the "we need 12 months to integrate our value add".... is kind of trigger words
20:38:07 cdent heh
20:38:37 cdent I have a sort of point of order question, or something like that:
20:39:35 cdent Why are people so worried about shit spiralling out of control? From a slightly different point of view that looks like people just avoiding the truth and putting more barriers in place to getting to the root causes of problems.
20:40:16 cdent I know, as mriedem so eloquently puts it, there's a lot of bitter feelings, but that's a symptom of a greater disease, not the problem itself

Earlier   Later