| xgerman | ok, so Aish has the right config but still SSL error? | 00:10 |
|---|---|---|
| xgerman | wrong window | 00:10 |
| *** manishg has joined #openstack-lbaas | 00:12 | |
| johnsom | https://review.openstack.org/#/c/288187 | 00:14 |
| johnsom | ^^^ CLI for cascading deletes. WIP until you folks can make a decision. | 00:14 |
| *** diogogmt has quit IRC | 00:28 | |
| blogan | dougwig: how will the get me a network cli look? | 00:30 |
| johnsom | oh, good question | 00:31 |
| xgerman | ugly | 00:31 |
| xgerman | I don;’t think the CLI should have that | 00:33 |
| xgerman | I would say we go with our original plan and cross that bridge when that tree-hugger stuff is baked out better | 00:36 |
| xgerman | then we can always deprecate the cascade delete endpoint and switch to tree | 00:36 |
| *** manishg has quit IRC | 00:38 | |
| dougwig | Nova has list show delete ... | 00:40 |
| xgerman | well, they also don;’t call their stuff tree | 00:45 |
| xgerman | or vm-tree | 00:45 |
| *** manishg has joined #openstack-lbaas | 00:46 | |
| dougwig | Lbaas-create doesn't dontain tree. :) | 00:46 |
| xgerman | neither does their API | 00:50 |
| jwarendt | Added comment about the '--recursive' flag used by containers for delete, with unfortunate client-side semantics, to michael's patch. | 00:50 |
| xgerman | yeah, I like recursive | 00:56 |
| jwarendt | At the REST boundary, I'm for a separate resource but following convention: ../lb_trees Plural, is a collection. The instance is ../lb_trees/{id} | 00:57 |
| johnsom | jwarendt So what does GET /lb_trees mean? | 00:57 |
| jwarendt | It should mean "Get the representation of a set of lb_trees instances". Just like pools or healthmonitors or loadbalancers. | 00:58 |
| jwarendt | Can be empty. | 00:58 |
| johnsom | But a tree is not an object here. pools, healthmonitors, loadbalancers are | 00:59 |
| xgerman | and hence I dislike tree | 00:59 |
| xgerman | as much as I like bike shedding as the next guy | 01:02 |
| xgerman | w need to come up with some endpoint for that cascade delete | 01:02 |
| xgerman | I don;t think we know enough about status_tree, get-me-an-LB, and such to come up with how that will look in REST | 01:03 |
| xgerman | so we should just pick something which makes sense for cascade_delete | 01:03 |
| xgerman | without putting all that stuff into consideration | 01:03 |
| johnsom | I need to step away for a bit. What I need from the team is: API path, lbaas-loadbalancer-cascade-delete or recursive flag on current delete call. | 01:03 |
| johnsom | No answer by 8am tomorrow morning and I will take the WIP off and push for merge as is. | 01:04 |
| xgerman | well, my answer has been known for days | 01:04 |
| johnsom | Just because we need to get this stuff in for Mitaka | 01:05 |
| sbalukoff | Just sent a long e-mail to the openstack-dev list complaining about documentation woes. | 01:05 |
| xgerman | tahnks | 01:05 |
| xgerman | you missed an afternoon of bike shedding | 01:05 |
| johnsom | sbalukoff FYI, they have a separate docs mailing list | 01:05 |
| sbalukoff | Are you kidding me? | 01:06 |
| sbalukoff | Argh! | 01:06 |
| johnsom | Nope, not kidding | 01:06 |
| sbalukoff | It's like they really don't want anyone contributing documentation. | 01:06 |
| johnsom | But thanks for the e-mail. I will read it later | 01:06 |
| sbalukoff | *sigh* | 01:06 |
| xgerman | OpenStack is all about crushing dreams | 01:06 |
| sbalukoff | Ok, so what is this list to which I should forward the long e-mail I just wrote? | 01:06 |
| sbalukoff | xgerman: Apparently! | 01:06 |
| *** Paco_ has joined #openstack-lbaas | 01:07 | |
| *** piet has quit IRC | 01:07 | |
| sbalukoff | Also, johnsom: I see you sent an e-mail about our documentation disappearing mysteriously from the Mitaka manual. Thanks for that! | 01:07 |
| xgerman | +1 | 01:08 |
| openstackgerrit | min wang proposed openstack/octavia: WIP----Replace the os.open method with safer way https://review.openstack.org/288208 | 01:09 |
| *** minwang2 has quit IRC | 01:12 | |
| blogan | xgerman: we already have status tree | 01:12 |
| *** blogan is now known as blogan-bot | 01:14 | |
| *** blogan-bot is now known as blogan | 01:14 | |
| *** Aish has left #openstack-lbaas | 01:21 | |
| *** ducttape_ has joined #openstack-lbaas | 01:24 | |
| *** ducttape_ has quit IRC | 01:25 | |
| *** ducttape_ has joined #openstack-lbaas | 01:25 | |
| *** chenli has joined #openstack-lbaas | 01:26 | |
| *** ducttape_ has quit IRC | 01:31 | |
| *** diogogmt has joined #openstack-lbaas | 01:35 | |
| openstackgerrit | Henry Gessau proposed openstack/neutron-lbaas: Tag the alembic migration revisions for Mitaka https://review.openstack.org/288214 | 01:38 |
| *** madhu_ak has quit IRC | 01:46 | |
| *** yamamoto has joined #openstack-lbaas | 01:53 | |
| *** neatherweb has joined #openstack-lbaas | 01:58 | |
| *** yuanying has quit IRC | 01:58 | |
| *** yuanying has joined #openstack-lbaas | 01:59 | |
| *** clduser has quit IRC | 02:03 | |
| *** bana_k has quit IRC | 02:17 | |
| yamamoto | any chance this can be merged quickly? https://review.openstack.org/#/c/287014/ | 02:19 |
| *** fawadkhaliq has quit IRC | 02:20 | |
| yamamoto | this is blocking midonet gate. we can workaround by folding unrelated patches but i want to avoit it if possible. | 02:20 |
| *** yamamoto has quit IRC | 02:23 | |
| *** yamamoto has joined #openstack-lbaas | 02:23 | |
| *** ducttape_ has joined #openstack-lbaas | 02:25 | |
| xgerman | blogan: LBaaS/load balancers/<id>/statuses this is making the argument for /cascade_delete | 02:27 |
| blogan | i know | 02:27 |
| xgerman | 😃 | 02:29 |
| *** yamamoto has quit IRC | 02:29 | |
| *** clduser has joined #openstack-lbaas | 02:32 | |
| *** woodster_ has quit IRC | 02:47 | |
| *** bana_k has joined #openstack-lbaas | 02:49 | |
| sbalukoff | yamamoto: I would if I could. :/ | 02:53 |
| sbalukoff | Oh, he left. | 02:53 |
| *** piet has joined #openstack-lbaas | 02:53 | |
| *** links has joined #openstack-lbaas | 03:01 | |
| *** yamamoto has joined #openstack-lbaas | 03:11 | |
| *** piet has quit IRC | 03:13 | |
| *** piet has joined #openstack-lbaas | 03:14 | |
| *** yuanying has quit IRC | 03:16 | |
| *** bana_k has quit IRC | 03:20 | |
| *** ducttape_ has quit IRC | 03:36 | |
| *** fawadkhaliq has joined #openstack-lbaas | 03:38 | |
| *** ducttape_ has joined #openstack-lbaas | 03:42 | |
| *** links has quit IRC | 03:44 | |
| *** minwang2 has joined #openstack-lbaas | 03:46 | |
| *** Purandar has quit IRC | 03:50 | |
| *** kevo has quit IRC | 03:51 | |
| johnsom | sbalukoff It's at http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-docs | 03:54 |
| sbalukoff | johnsom: Thanks! | 03:54 |
| sbalukoff | johnsom: I've actually got a conversation going with someone who obviously knows a lot more about the doc system than we do... | 03:54 |
| sbalukoff | It's been immensely helpful. So, look for my response on that momentarily. | 03:55 |
| sbalukoff | Also, I'm pretty sure I pissed off armax. | 03:55 |
| sbalukoff | ;) | 03:55 |
| johnsom | Well, that is good. After talking with people off and on all day and sending an e-mail as requested, I have got zero response. | 03:55 |
| *** ducttape_ has quit IRC | 03:56 | |
| johnsom | Well, that is quite the e-mail. | 03:56 |
| johnsom | I will say, part of the reason this came up is *because* we put in the effort and care, but found them all deleted! | 03:57 |
| sbalukoff | Haha! I know! | 03:57 |
| sbalukoff | Lana's response to me on the doc list is actually quite good. I'm composing a reply now with some suggestions... | 03:58 |
| sbalukoff | But in any case, I think she answered enough of my questions so that I can get started on fixing docs... | 03:58 |
| sbalukoff | We'll still have to find out why the heck our docs disappeared as I think you are right that something we don't control is broken there. | 03:59 |
| johnsom | Yeah, I found out why, it was in my e-mail. They deleted "third party driver docs". | 04:00 |
| johnsom | My big problem is here is a team that put a bunch of effort to put those in, went to update them, and found them deleted. | 04:00 |
| sbalukoff | It may turn out to be easier for us to document this stuff in the neutron-lbaas repository the same way we do in the Octavia repository... but that doesn't help much with an end-user or operator guide. :/ | 04:03 |
| sbalukoff | (It turns out that that's not necessarily considered a bad thing to do it that way...) | 04:03 |
| sbalukoff | Unless, of course, neutron leadership disagrees with us. :P | 04:03 |
| johnsom | I guess, so but how does that link in to the existing docs? | 04:07 |
| *** yuanying has joined #openstack-lbaas | 04:08 | |
| johnsom | I mean the only reason I would make that kind of a change is to stop people from deleting docs | 04:08 |
| *** yuanying has quit IRC | 04:12 | |
| sbalukoff | Heh! True enough. | 04:13 |
| sbalukoff | In any case, there are not good guidelines right now on what should go in the openstack manual, and what should go in project-specific documentation within the project repo. | 04:14 |
| *** chenli has quit IRC | 04:15 | |
| *** Purandar has joined #openstack-lbaas | 04:17 | |
| *** numans has joined #openstack-lbaas | 04:17 | |
| *** piet has quit IRC | 04:27 | |
| *** links has joined #openstack-lbaas | 04:27 | |
| *** openstack has joined #openstack-lbaas | 14:09 | |
| openstackgerrit | Ihar Hrachyshka proposed openstack/neutron-lbaas: Make all tox targets constrained https://review.openstack.org/288431 | 14:10 |
| *** neelashah has joined #openstack-lbaas | 14:14 | |
| *** piet has quit IRC | 14:16 | |
| *** yamamoto_ has quit IRC | 14:17 | |
| openstackgerrit | Ihar Hrachyshka proposed openstack/neutron-lbaas: tox: constrain releasenotes and cover targets https://review.openstack.org/288472 | 14:18 |
| *** amotoki has quit IRC | 14:21 | |
| *** ducttape_ has joined #openstack-lbaas | 14:21 | |
| *** ducttape_ has quit IRC | 14:26 | |
| *** piet has joined #openstack-lbaas | 14:36 | |
| *** openstackgerrit_ has quit IRC | 14:36 | |
| *** openstackgerrit_ has joined #openstack-lbaas | 14:38 | |
| *** links has quit IRC | 14:41 | |
| *** diogogmt has quit IRC | 14:42 | |
| *** anilvenkata has quit IRC | 14:53 | |
| *** ducttape_ has joined #openstack-lbaas | 14:56 | |
| *** piet has quit IRC | 14:56 | |
| *** diogogmt has joined #openstack-lbaas | 14:57 | |
| *** woodster_ has joined #openstack-lbaas | 15:03 | |
| *** piet has joined #openstack-lbaas | 15:04 | |
| *** yamamoto has joined #openstack-lbaas | 15:05 | |
| *** yamamoto has quit IRC | 15:05 | |
| *** yamamoto has joined #openstack-lbaas | 15:06 | |
| *** piet has quit IRC | 15:21 | |
| *** manishg has joined #openstack-lbaas | 15:33 | |
| *** yamamoto has quit IRC | 15:39 | |
| *** Bjoern has joined #openstack-lbaas | 15:42 | |
| *** numans has quit IRC | 15:47 | |
| *** kiran-r has quit IRC | 15:51 | |
| *** kiran-r has joined #openstack-lbaas | 15:53 | |
| *** yamamoto has joined #openstack-lbaas | 15:55 | |
| *** manishg has quit IRC | 15:58 | |
| *** kobis has joined #openstack-lbaas | 16:01 | |
| *** yamamoto has quit IRC | 16:05 | |
| *** TrevorV has joined #openstack-lbaas | 16:17 | |
| *** Purandar has joined #openstack-lbaas | 16:18 | |
| *** piet has joined #openstack-lbaas | 16:27 | |
| openstackgerrit | Victor Stinner proposed openstack/neutron-lbaas: Fix Python 3 issues https://review.openstack.org/288382 | 16:28 |
| openstackgerrit | Victor Stinner proposed openstack/neutron-lbaas: Port test_agent and test_agent_api to Python 3 https://review.openstack.org/288414 | 16:28 |
| openstackgerrit | Victor Stinner proposed openstack/neutron-lbaas: Add a whitelist of tests passing on Python 3 https://review.openstack.org/288412 | 16:28 |
| openstackgerrit | Victor Stinner proposed openstack/neutron-lbaas: Port cert_parser to Python 3 https://review.openstack.org/288413 | 16:28 |
| openstackgerrit | Victor Stinner proposed openstack/neutron-lbaas: py3: Use six.moves for import modules https://review.openstack.org/288410 | 16:28 |
| openstackgerrit | Victor Stinner proposed openstack/neutron-lbaas: py3: Replace dict.iteritems() with dict.items() https://review.openstack.org/288411 | 16:28 |
| *** armax has quit IRC | 16:29 | |
| *** armax has joined #openstack-lbaas | 16:29 | |
| *** jcook_ has joined #openstack-lbaas | 16:29 | |
| *** kobis has quit IRC | 16:40 | |
| *** rcernin has quit IRC | 16:51 | |
| *** manishg has joined #openstack-lbaas | 16:51 | |
| *** ihrachys has quit IRC | 16:58 | |
| *** Aish has joined #openstack-lbaas | 16:58 | |
| *** kevo has quit IRC | 17:00 | |
| *** amotoki has joined #openstack-lbaas | 17:03 | |
| *** yamamoto has joined #openstack-lbaas | 17:05 | |
| *** minwang2 has joined #openstack-lbaas | 17:06 | |
| *** kiran-r has quit IRC | 17:08 | |
| openstackgerrit | Merged openstack/octavia: Updated from global requirements https://review.openstack.org/288358 | 17:11 |
| TrevorV | xgerman dougwig johnsom anyone around that I can talk to for a sec? | 17:13 |
| johnsom | I am in a bug scrub call, but can chat on IRC | 17:13 |
| TrevorV | Awesome | 17:14 |
| TrevorV | So as we know i'm working on the single-create | 17:14 |
| TrevorV | I'm not making any progress on "modifying" the current "POST LB" | 17:14 |
| TrevorV | Would you be wildly against single-create being a new endpoint/resource in neutron lbaas? | 17:14 |
| *** yamamoto has quit IRC | 17:15 | |
| johnsom | I think we were talking about that yesterday. Putting cascade delete and maybe even create under /lbaas/lb_tree/<lbid> or something similar | 17:15 |
| johnsom | blogan Did they bike shredder's make a call? | 17:16 |
| TrevorV | johnsom he's not quite in the office yet | 17:16 |
| johnsom | So, something like that would make your life easier? | 17:17 |
| TrevorV | I think so, yeah. Though, I would like to note that I don't necessarily think its the *right* thing to do | 17:19 |
| johnsom | Oh no, more painting... I'll bite, what do you think is the *right* answer? | 17:20 |
| *** bana_k has joined #openstack-lbaas | 17:21 | |
| TrevorV | ha ha ha, personally it feels funky to have another POST endpoint that creates the same object that another endpoint does. | 17:21 |
| TrevorV | Meaning, it honestly should be added functionality to the POST LB endpoint | 17:22 |
| TrevorV | However I don't feel strongly enough about it that I wouldn't make a new endpoint (as suggested above), just saying if I had written the API on my own from ground up, that's how I would do it, you know? | 17:22 |
| johnsom | Yep | 17:23 |
| blogan | im here | 17:23 |
| TrevorV | blogan do you have scrollback? | 17:23 |
| johnsom | blogan Welcome, grab a brush... So, did we come to some conclusion on the API path? | 17:23 |
| blogan | one of the problems with the lb_tree endpoint is that we'll need to also implement the get_lb_trees, get_lb_tree, delete_lb_tree, and create_lb_tree to make it correct. If we just do cascade delete for M, it'll be quite weird to only have the delete_lb_tree working | 17:24 |
| blogan | johnsom: we did not | 17:24 |
| johnsom | SOG | 17:25 |
| johnsom | SOB | 17:25 |
| xgerman | well, I reiterated my arguments and nobody objected | 17:25 |
| *** Aish has left #openstack-lbaas | 17:25 | |
| xgerman | agree with TrevorV we are creating a load balancer so it should go on it’s endpoint | 17:25 |
| xgerman | maybe create LB can be a social case of LB_tree — don’t we send JSOPN anyway? | 17:25 |
| johnsom | We don't send json with delete | 17:26 |
| xgerman | after all creating a LB is like creating an LB tree without listeners, members, and pools | 17:26 |
| johnsom | I kind of agree with that | 17:26 |
| xgerman | delete should go on loadbalancer/<id>/cascade_delete | 17:27 |
| xgerman | TrevorV’s create on loadbalancer | 17:27 |
| xgerman | status tree is on loadbalancer/<id>/statuses | 17:27 |
| xgerman | problem solved — putting down my brush | 17:27 |
| TrevorV | See that may be another problem though, xgerman | 17:28 |
| TrevorV | The change I'd make (that I'd have to figure out how it works exactly) would be to just accept any given object under the load balancer. Meaning it isn't extended functionality, its modifying original functionality | 17:28 |
| TrevorV | However we can't do that because of the other extensions | 17:29 |
| TrevorV | I forget the word | 17:29 |
| TrevorV | I don't mean extensions | 17:29 |
| TrevorV | I mean the... "providers"? The competitors to "octavia" shit, words. | 17:29 |
| TrevorV | Drivers | 17:29 |
| TrevorV | That's the word | 17:29 |
| TrevorV | If we change POST LB the way I think it should be changed, all the drivers would need changed, rather than "optionally" changing | 17:30 |
| rm_work | i was kinda liking the feeling of GET/PUSH/DELETE on /tree/ for getting the status tree, doing a full create, and doing cascade delete | 17:30 |
| rm_work | err | 17:30 |
| rm_work | GET/POST/DELETE | 17:30 |
| rm_work | it's early <_< | 17:30 |
| rm_work | obviously /tree/ is a working-title for the endpoint <_< | 17:31 |
| xgerman | if we talk verbs — it seems we need PATCH as well for a partial tree update IHMO | 17:31 |
| *** Aish has joined #openstack-lbaas | 17:31 | |
| TrevorV | xgerman I don't follow that argument. single-create should support whatever valid object is given, even if it is only an LB and a Listener | 17:32 |
| xgerman | we are creating a load balancer so it should be on loadbalancer | 17:32 |
| xgerman | I get your argument that your create format is not backwards compatible with the single create | 17:33 |
| *** lunarlamp has joined #openstack-lbaas | 17:33 | |
| *** Aish has left #openstack-lbaas | 17:33 | |
| openstackgerrit | Stephen Balukoff proposed openstack/octavia: Make SSH bind to management net only https://review.openstack.org/288323 | 17:33 |
| xgerman | but I am wondering if there is a way to detect which format is being passed in — or would that then get too clunky | 17:34 |
| xgerman | ? | 17:34 |
| TrevorV | xgerman blogan would be better to answer that one, at least in concept | 17:34 |
| blogan | the single create call is additive, so it really is backwards incompatible | 17:35 |
| xgerman | ok, what I am proposing is the Content-Negotiation/Microversioning way of things | 17:35 |
| blogan | unless someone, something, was relying on the API not allowing listeners to be passed | 17:35 |
| TrevorV | xgerman I'm sorry I don't know what that means | 17:37 |
| *** kevo has joined #openstack-lbaas | 17:38 | |
| johnsom | FYI, place to see rendered pre-release release notes: http://docs.openstack.org/releasenotes/octavia/unreleased.html | 17:42 |
| johnsom | http://docs.openstack.org/releasenotes/neutron-lbaas/unreleased.html | 17:42 |
| TrevorV | Alright, I'll brb, gonna pick up lunch and come back | 17:44 |
| TrevorV | Maybe half hour and we can pick up this convo? | 17:44 |
| *** fnaval has joined #openstack-lbaas | 17:47 | |
| johnsom | Also, blogan rm_work we have a number of patches hanging out with +2's in Octavia. It would be nice to start getting these reviewed and merged. | 17:49 |
| *** ErnieP has joined #openstack-lbaas | 17:49 | |
| rm_work | ok can try to get to those in a few | 17:50 |
| johnsom | Thanks | 17:50 |
| rm_work | stuck on firefighter duty ATM | 17:50 |
| ErnieP | Hi All - is there a good doc for how to startup and configure octavia? | 17:51 |
| *** yamamoto has joined #openstack-lbaas | 17:53 | |
| blogan | ErnieP: https://review.openstack.org/#/c/278128/ | 17:54 |
| openstackgerrit | Stephen Balukoff proposed openstack/octavia: Make SSH bind to management net only https://review.openstack.org/288323 | 17:54 |
| blogan | ErnieP: it's still in review, but its probably your best bet | 17:54 |
| blogan | johnsom: will try to carve out some time today | 17:54 |
| johnsom | Thanks | 17:54 |
| blogan | my schedule is shrinking as we speak :( | 17:54 |
| ErnieP | blogan: thanks I will check it out | 17:55 |
| johnsom | ErnieP There is also the liberty guide for octavia.conf: http://docs.openstack.org/liberty/config-reference/content/networking-plugin-lbaas.html | 17:55 |
| blogan | ErnieP: np, come back if you have questions, i dont want to RTFM you, but that covers a good amount | 17:55 |
| johnsom | ErnieP Sadly we are having some issues with the docs/manuals repo at the moment so the Mitaka version is missing | 17:56 |
| *** bana_k has quit IRC | 17:57 | |
| ErnieP | I am happy to RTFM :-) I just couldn't find a good doc. The liberty doc just says config the plugin.. it says nothing about how to config octavia. | 17:57 |
| *** madhu_ak has joined #openstack-lbaas | 17:57 | |
| ErnieP | I did pip install octavia and it didn't get a /etc/octavia/octavia.conf file so I feel like there is a bootstrap step that I missed once the python libs are instaled | 17:58 |
| blogan | ErnieP: yeah some docs magically disappeared it was johnsom was alluding to | 17:58 |
| sbalukoff | ErnieP: We're going to be overhauling a lot of those docs in the next week as well. Apologies that we don't have them ready for your consumption yet. | 17:59 |
| blogan | ErnieP: whoops that review doesn't include much about octavia, sorry | 17:59 |
| sbalukoff | BBIAB | 17:59 |
| fnaval | review requested please: https://review.openstack.org/#/c/172199 | 18:00 |
| madhu_ak | likewise, need reviews for https://review.openstack.org/#/c/284946/, so the project config patch can be tested: https://review.openstack.org/#/c/284875/ | 18:01 |
| ErnieP | blogan: What do you need to do once the octavia python libs are installed? I don't need a step-by-step by high-level overivew would get me running. | 18:02 |
| blogan | ErnieP: I've honestly not installed it from packages, but I'm assuming it just creates the executables | 18:03 |
| blogan | octavia-api, octavia-cw | 18:03 |
| ErnieP | blogan: are those pypi packages | 18:03 |
| blogan | pretty sure octavia is one pypi package that will include those on install | 18:04 |
| blogan | but johnsom did the pypi stuff | 18:04 |
| ErnieP | blogan: would it just be better to do this from source ? | 18:04 |
| blogan | ErnieP: thats how i do it, just python setup develop (or install prefer) | 18:05 |
| johnsom | It is one pypi package we released for liberty. | 18:05 |
| johnsom | ErnieP The devstack setup script we use might help you with the steps: https://github.com/openstack/octavia/blob/master/devstack/plugin.sh | 18:05 |
| blogan | i was just about ot link that | 18:06 |
| blogan | ErnieP: it's for all-in-one though | 18:06 |
| johnsom | Right, that is just an example. | 18:06 |
| blogan | ok gotta go to lunch and then meetings after | 18:08 |
| blogan | ErnieP: hopefull you can decipher bash (i hate bash), if not i and others will be on throughout the dya | 18:09 |
| openstackgerrit | Michael Johnson proposed openstack/octavia: Adds release notes for Active/Standby https://review.openstack.org/288620 | 18:10 |
| ErnieP | blogan: and johnsom: thanks let me dig into it more | 18:13 |
| ErnieP | Is anyone using the haproxy plugin with v2 instead of octavia | 18:13 |
| *** yamamoto has quit IRC | 18:14 | |
| *** bana_k has joined #openstack-lbaas | 18:15 | |
| *** yamamoto has joined #openstack-lbaas | 18:15 | |
| *** Purandar has quit IRC | 18:16 | |
| *** Purandar has joined #openstack-lbaas | 18:18 | |
| *** piet has quit IRC | 18:19 | |
| *** piet has joined #openstack-lbaas | 18:21 | |
| *** yamamoto has quit IRC | 18:25 | |
| *** ducttape_ has quit IRC | 18:26 | |
| *** amotoki has quit IRC | 18:28 | |
| *** shakamunyi has joined #openstack-lbaas | 18:29 | |
| openstackgerrit | Merged openstack/octavia: Add bandit baseline to tox https://review.openstack.org/286892 | 18:31 |
| *** kobis has joined #openstack-lbaas | 18:35 | |
| sbalukoff | I'm going to be sporadically available throughout the day as I deal with some internal stuff. | 18:35 |
| sbalukoff | For example, I'll be AFK for at least an hour starting now. | 18:35 |
| sbalukoff | When I get back: johnsom: Let me know if you want me to dive into writing docs or back-filling those L7 unit tests first! | 18:36 |
| johnsom | sbalukoff Sorry, was lighting a fire: https://review.openstack.org/#/c/288630 | 18:37 |
| *** Purandar has quit IRC | 18:38 | |
| *** ducttape_ has joined #openstack-lbaas | 18:38 | |
| *** piet has quit IRC | 18:38 | |
| *** kobis has quit IRC | 18:40 | |
| *** sbalukoff has quit IRC | 18:40 | |
| *** Purandar has joined #openstack-lbaas | 18:40 | |
| *** pcaruana has quit IRC | 18:41 | |
| *** piet has joined #openstack-lbaas | 18:41 | |
| *** Aish has joined #openstack-lbaas | 18:49 | |
| TrevorV | Okay sorry guys, my lunch retrieval went a lot longer than it should have/ | 18:49 |
| TrevorV | So if I can steal some time from everyone to actually reach an agreement here so I'm not idly waiting to make changes, that'd be super awesome | 18:50 |
| *** doug-fis_ is now known as doug-fish | 18:50 | |
| *** Purandar has quit IRC | 18:50 | |
| TrevorV | johnsom rm_work blogan xgerman and anyone else interested or expressing concern | 18:50 |
| TrevorV | dougwig maybe has an opinion to share too | 18:51 |
| johnsom | Yeah, I would really like a decision too | 18:51 |
| xgerman | I have 9 minutes until my next meeting ;-) | 18:51 |
| * dougwig raises his head. | 18:51 | |
| TrevorV | xgerman maybe table this until you're available and my team is back from lunch? ha ha | 18:51 |
| johnsom | Yeah, sbalukoff just left too | 18:51 |
| dougwig | crud /trees420/ is my vote. | 18:52 |
| TrevorV | I'll put it this way, internally Rax needs single-create available by next friday | 18:52 |
| TrevorV | So I need to make something happen | 18:52 |
| TrevorV | (I don't mean merged upstream next Friday, but we need a review available for consumption by then) | 18:52 |
| johnsom | We need cascading delete before then | 18:52 |
| xgerman | dougwig I would expect that from Idaho — I like CRUD /cactus/ | 18:52 |
| dougwig | from idaho? like, the last state that will ever legalize weed? | 18:53 |
| xgerman | wanna do CRUD /weed/ ? | 18:53 |
| *** Purandar has joined #openstack-lbaas | 18:53 | |
| TrevorV | So if we're going to address single-create in the same concept as cascade delete, then I would say the correct behavior is modifying delete behavior the same way we should modify create behavior | 18:54 |
| xgerman | well, since the Get-me-an-LB gets a loadbalancer it should be under that endpoint | 18:54 |
| TrevorV | xgerman so you're saying it needs a new endpoint? | 18:55 |
| xgerman | I am saying use the existing endpoint and inure out based on what you pass it what should happen | 18:55 |
| xgerman | figure | 18:55 |
| xgerman | also micro versioning is supposed to solve that — so will Get-me-an-LB land BEFORE we have that? | 18:56 |
| TrevorV | xgerman so I don't follow what that is, the micro versioning thing... Does neutron lbaas have that? | 18:58 |
| xgerman | neutron is supposed to get that | 18:58 |
| xgerman | then you can have different behavior based on the version on the same endpoint | 18:59 |
| TrevorV | So it has it or it doesn't? That's what I'm asking | 18:59 |
| TrevorV | sorry, "supposed to get that" means to me that it isn't ready in Neutron, but should be soon | 19:00 |
| TrevorV | What is soon in this case? | 19:00 |
| TrevorV | Oh, yeah, he just disappeared to a meeting | 19:02 |
| TrevorV | Alright, we'll table this until we have more hands on deck I guess | 19:02 |
| TrevorV | dougwig I'll keep you marked down as the "new endpoint" opinion, is that accurate? | 19:02 |
| *** ihrachys has joined #openstack-lbaas | 19:10 | |
| *** yamamoto has joined #openstack-lbaas | 19:16 | |
| *** yamamoto has quit IRC | 19:21 | |
| openstackgerrit | Jiahao Liang proposed openstack/octavia: Update doc to make it consistent with the actual api. https://review.openstack.org/288674 | 19:23 |
| *** kev0 has joined #openstack-lbaas | 19:24 | |
| *** kevo has quit IRC | 19:24 | |
| dougwig | TrevorV: aye | 19:29 |
| *** piet has quit IRC | 19:35 | |
| *** piet has joined #openstack-lbaas | 19:36 | |
| *** ihrachys has quit IRC | 19:38 | |
| blogan | microversioning is not something we should count on anytime soon | 19:49 |
| *** sbalukoff has joined #openstack-lbaas | 19:50 | |
| TrevorV | okay, so that answers my question, totally not available near future | 19:51 |
| TrevorV | got it | 19:51 |
| blogan | so if all calls for lb_tree were ready today, i'd go with lb_tree | 19:51 |
| blogan | or even if all lb_tree calls could be postponed until N | 19:52 |
| *** fawadkhaliq has joined #openstack-lbaas | 19:52 | |
| blogan | but since we NEED the delete lb_tree functionality, that's just going to make the extensions wonky to do cascade delete first as one extension (an extension that will be wonky in itself for lb_tree) and then fill otu the rest of the lb tree functionality in N | 19:53 |
| TrevorV | blogan you lost me there, you're saying new endpoint? | 19:54 |
| blogan | no, new endpoint will make things wonky since we need DELETE in now | 19:54 |
| TrevorV | so since micro-versioning isn't a thing right now, we're talking changing POST/DELETE behavior to just do what we want and hope we don't break drivers? | 19:55 |
| blogan | no delete would be a new resource off the loadbalancer | 19:56 |
| blogan | /loadbalancer/id/cascade_delete | 19:56 |
| blogan | what's currently in review | 19:56 |
| blogan | vs DELETE /lb_tree/id | 19:57 |
| TrevorV | Oooh you and I are talking about the same thing, but I keep using the wrong jargon. | 19:58 |
| TrevorV | So you're in favor of new resource, but are you in favor of new resource for single-create as well? Or should that be a new endpoint? | 19:58 |
| *** bhaargavi has joined #openstack-lbaas | 19:59 | |
| TrevorV | blogan ^^? | 19:59 |
| blogan | if we do /loadbalancer/id/casccade_delete then keep create as it is | 20:01 |
| blogan | keep it how the single create already has it | 20:02 |
| blogan | in its current review form | 20:02 |
| TrevorV | As an extension. Alright, so, in your opinion, if we say cascade delete needs to be off a new endpoint, DELETE on "/lb_graph/<id>" or whatnot, then the single-create should be a POST on that same endpoint | 20:04 |
| TrevorV | Otherwise all is well and should be iterated upon | 20:04 |
| blogan | yes, it'll be an extension either way | 20:04 |
| *** piet has quit IRC | 20:11 | |
| *** yamamoto has joined #openstack-lbaas | 20:16 | |
| *** yamamoto has quit IRC | 20:21 | |
| TrevorV | xgerman johnsom we ready to continue? | 20:22 |
| *** piet has joined #openstack-lbaas | 20:25 | |
| *** yamamoto has joined #openstack-lbaas | 20:37 | |
| *** yamamoto has quit IRC | 20:37 | |
| *** yamamoto has joined #openstack-lbaas | 20:39 | |
| *** yamamoto has quit IRC | 20:39 | |
| *** piet has quit IRC | 21:07 | |
| sbalukoff | Our py34 broke spontaneously? | 21:07 |
| sbalukoff | The check, I mean. | 21:08 |
| xgerman | blogan how’s micro versioning coming; can we add the Get-me-an-LB on /loadbalancer? | 21:11 |
| xgerman | I really feel having a second endpoint which returns an LB clunky | 21:11 |
| xgerman | but I am also worn down and like to close… it’s Friday afternoon after all | 21:12 |
| blogan | xgerman: microversioning not going to happen anytime soon in neutron | 21:12 |
| xgerman | boo | 21:12 |
| xgerman | how do they fell about “content-negotiation” — you look at the body and figure out what should happen? | 21:12 |
| xgerman | after all creating one LB is a special case of creating an LB, a listener, etc. | 21:13 |
| xgerman | signal create is a special case of get-me-an-LB | 21:13 |
| blogan | i'm sure they don't feel strong one or the other to allow it to get in before M :) | 21:17 |
| xgerman | well, I guess you get my drift. Post single create to /loadbalancer -> single create | 21:18 |
| xgerman | Post Get-me-an-LB content to /loadbalancer -> get-me-an-LB | 21:19 |
| xgerman | future make it so that they look more or less the same | 21:19 |
| xgerman | but if we do the wrong thing we can always fix it in LBaaS 3.0 | 21:20 |
| TrevorV | xgerman what are you talking about? "get-me-an-LB" isn't the same as "single-create"? | 21:20 |
| xgerman | so if I do get-me-an-LB and only define an LB — do I get back an LB? | 21:21 |
| blogan | lbaas 3.0 will hopefully be octavia's api | 21:21 |
| xgerman | so they have the same output... | 21:21 |
| rm_work | ^^ | 21:21 |
| xgerman | then labs 4.0 | 21:21 |
| TrevorV | xgerman I don't understand, what is "get-me-an-lb"? Does that exist right now? | 21:21 |
| blogan | hopefully we get correct versioning design in octavia though | 21:22 |
| xgerman | +1 | 21:22 |
| blogan | get me an lb is single create | 21:22 |
| xgerman | ah | 21:22 |
| xgerman | hoe do you call what e have right now? | 21:22 |
| blogan | xgerman: right now as in whats in master? or whats upf or review? | 21:23 |
| xgerman | well, may regiment is that single create returns an LB like what is lurking right now behind /loadbalancer | 21:25 |
| blogan | yes | 21:25 |
| blogan | thats how it is in reveiw | 21:25 |
| xgerman | so /loadbalancer is a special case of single create | 21:25 |
| xgerman | hence single create should post to laodbalancer | 21:25 |
| blogan | the way it is right now POST /loadbalancer with a body that includes listeners will create everything and return everything that got created in the return body | 21:25 |
| xgerman | yeah, that makes a ton of sense | 21:26 |
| xgerman | so keep it | 21:26 |
| xgerman | +2 | 21:26 |
| TrevorV | Except that breaks drivers potentially xgerman I believe | 21:26 |
| blogan | TrevorV: tahts why there's a driver check to see if it supports it | 21:26 |
| xgerman | ok, so we are good? | 21:26 |
| blogan | TrevorV: if it doesn't, its not different than if we implemented lb_tree and the driver didn't support it | 21:27 |
| xgerman | +1 | 21:27 |
| TrevorV | Alright, if that's the right way, then alrighty, all good here. That leaves dougwig sad though | 21:27 |
| xgerman | just send him a pound of weed in the mail | 21:28 |
| TrevorV | owning 2 ounces is a felony.... so that won't happen :P | 21:28 |
| xgerman | just start a dispensary ;-) | 21:28 |
| rm_work | alright johnsom i just did a sweep of reviews | 21:30 |
| rm_work | johnsom: there's a few more I +2'd that need another as well | 21:30 |
| rm_work | and had a question on one of sbalukoff's | 21:31 |
| rm_work | question/nit | 21:31 |
| rm_work | always feel guilty doing nitpicks on things with +2 already but <_< | 21:31 |
| rm_work | it's sbalukoff, he can take it | 21:31 |
| sbalukoff | ;_; | 21:32 |
| sbalukoff | Which one, rm_work? | 21:32 |
| sbalukoff | Also, isn't the py34 gate busted right now? | 21:32 |
| sbalukoff | You should also throw a +2 at this as it's been there for a while with one +2 and no comment: https://review.openstack.org/#/c/285143/ | 21:33 |
| sbalukoff | Ok, just saw: Yes, that's a good point. I'll fix it momentarily. | 21:35 |
| sbalukoff | (And I'm one who usually loathes magic numbers in code...) | 21:35 |
| dougwig | no way should single create post to an existing different object create. bleh. | 21:35 |
| dougwig | i get why that feels right, but we fucked it up and then shipped it, so we don't get a do-over. | 21:36 |
| dougwig | IMO | 21:36 |
| TrevorV | Okay, so we don't have agreement? | 21:37 |
| dougwig | don't let yourself be stalemated, because we won't get consensus here in a few days. you're going to have to make some people unhappy. | 21:38 |
| blogan | TrevorV: the extension will be totally different for this lb_tree endpoint, but the code in the plugin afterwards will be very similar, just may have to move the logic to anotehr method, not a big deal | 21:39 |
| *** yamamoto has joined #openstack-lbaas | 21:39 | |
| dougwig | just don't break backwards compat, and then yeah, we can always alias shim as you say. | 21:40 |
| dougwig | propose a non-additive semantic change, though, and that will never see the light of day. | 21:41 |
| TrevorV | dougwig I feel differently about that last bit, but I'll do my best. | 21:41 |
| rm_work | sbalukoff: one question on that one as well | 21:42 |
| blogan | dougwig: isn't accepting a full tree of a load balancer in the same request body as the old additive? | 21:42 |
| sbalukoff | rm_work: What is your question? | 21:43 |
| rm_work | i posted | 21:43 |
| sbalukoff | blogan: Because the same API will still accept the exact same old request? | 21:44 |
| sbalukoff | rm_work: Ok, resonded. | 21:47 |
| sbalukoff | responded. | 21:47 |
| sbalukoff | geez, I really can't type today. | 21:47 |
| sbalukoff | Is johnsom around? | 21:47 |
| *** yamamoto has quit IRC | 21:48 | |
| openstackgerrit | Stephen Balukoff proposed openstack/octavia: Better L7Policy API validations https://review.openstack.org/285783 | 21:49 |
| sbalukoff | rm_work: Ok, I updated this per your very good point ^^^ | 21:49 |
| blogan | sbalukoff: yes it'll accept the old one and more now! | 21:50 |
| blogan | thats additive | 21:50 |
| sbalukoff | blogan: Sounds good to me! | 21:50 |
| sbalukoff | I've doomed your side of the argument now, haven't I? | 21:50 |
| sbalukoff | Dang, I should have agreed with dougwig... | 21:51 |
| *** rtheis has quit IRC | 21:51 | |
| blogan | lol | 21:53 |
| blogan | well like i said, if we could get the entire lb_tree in (POST, DELETE, GETS) i'd be allf or it | 21:54 |
| blogan | for M | 21:54 |
| blogan | but we're only gonig to get delete in for M, and thats just going to be one extension, and then for N we're going to have to do another extension, and they're both going to be wonky | 21:54 |
| sbalukoff | Right. | 21:55 |
| sbalukoff | You really are preaching to the choir with me. :/ | 21:55 |
| blogan | so we're damned if we do damned if we dont. Which one are we less damned on though? | 21:56 |
| blogan | dougwig: how damned do you prefer to be? | 21:57 |
| *** pcaruana has joined #openstack-lbaas | 21:58 | |
| openstackgerrit | Stephen Balukoff proposed openstack/octavia: Make SSH bind to management net only https://review.openstack.org/288323 | 21:58 |
| sbalukoff | I can't reproduce that py34 gate check locally. And I notice the zuul status page isn't showing either of the patches I just uploaded... So... gate issues again today? | 22:01 |
| sbalukoff | Yeah, looks like they're troubleshooting it in the infra channel. | 22:02 |
| openstackgerrit | Madhusudhan Kandadai proposed openstack/octavia: Add pre_test_hook to run gate jobs https://review.openstack.org/284946 | 22:12 |
| openstackgerrit | Jiahao Liang proposed openstack/octavia: Update doc to make it consistent with the actual api https://review.openstack.org/288674 | 22:13 |
| *** fnaval has quit IRC | 22:17 | |
| johnsom | sbalukoff I am around now. | 22:17 |
| sbalukoff | johnsom: Oh, good! I just wanted to ask you how you'd like to proceed on the documentation side of things so we can hopefully not butt heads and / or duplicate effort. | 22:18 |
| bharathm | sbalukoff: py34 likely is a gate issue.. Your patch should be fine (I hope :-p) | 22:19 |
| sbalukoff | bharathm: yeah, they're talking about the gate problems in the infra irc channel. | 22:19 |
| johnsom | sbalukoff Feeling ok? Not butt heads???? | 22:19 |
| sbalukoff | johnsom: Haha! Well, documentation is hard enough as is, especially given so much of it often comes down to personal taste in writing style. ;) | 22:20 |
| sbalukoff | Unlike code where there's only ever one way to do things the right way. ;) | 22:20 |
| johnsom | Yeah, right | 22:20 |
| sbalukoff | Seriously though: It looks like there's an LBaaSv2 doc patch that's been getting attention over the last couple weeks. | 22:20 |
| johnsom | Well, since they have not responded in any useful and non-rude way in 24 hours I decided to poke them a bit by proposing a revert for the patch that deleted our docs | 22:21 |
| *** TrevorV has quit IRC | 22:21 | |
| sbalukoff | (With a ton of back-and-forth on what comes down to opinion in most cases... :P ) | 22:21 |
| sbalukoff | johnsom: Yeah, I saw that, and saw them shoot it down, too. :P | 22:21 |
| sbalukoff | Reeeeally annoying. | 22:21 |
| johnsom | Did they? Oh good | 22:21 |
| sbalukoff | I don't think they understood why you proposed the revert. | 22:22 |
| sbalukoff | At least, the comments I saw ignored the issue. | 22:22 |
| sbalukoff | Comments were basically, "we don't want to go back to xml! Are you nuts?") | 22:22 |
| rm_you | lol | 22:22 |
| rm_you | to be fair | 22:23 |
| johnsom | Well, that is a good response | 22:23 |
| rm_you | that is the definition of insanity | 22:23 |
| rm_you | I assume the point of proposing the revert wasn't that you actually expected to revert -- it was just to get them to pay attention, right? | 22:23 |
| sbalukoff | rm_you: Oh yes, I totally agree to the sentiment of the my paraphrased quote above. | 22:23 |
| johnsom | Sadly their advice was manually convert it, which I am now 99% convinced will just get removed again when the automatic process runs | 22:24 |
| johnsom | Basically it seems no one knows how this works now. | 22:25 |
| sbalukoff | johnsom: With that and armax's comments on the ML yesterday, I think maybe we should just dig in, decide where we want to go, propose changes to stuff that won't get auto-deleted, and be forceful about it. :/ | 22:25 |
| openstackgerrit | min wang proposed openstack/octavia: WIP----Replace the os.open method with safer way https://review.openstack.org/288208 | 22:25 |
| sbalukoff | johnsom: Yeah, basically. :/ | 22:25 |
| johnsom | Agreed. I did feel better seeing the number of conflicts in gerrit that raised however. | 22:26 |
| sbalukoff | johnsom: I felt like they were a lot more helpful on the -docs mailing list. Did you try e-mailing them there yet? | 22:26 |
| sbalukoff | johnsom: Yeah, that list is hillarious. :) | 22:27 |
| johnsom | I did, the morning before yours went out. No response | 22:27 |
| sbalukoff | Ok. | 22:27 |
| johnsom | Broke gerrit: 400 cannot create query for index: (path:RELEASENOTES.rst OR path:doc-tools-check-languages.conf OR | 22:27 |
| sbalukoff | johnsom: yep, gate issues right now. They're discussing in the infra channel. | 22:27 |
| johnsom | Anyway, so here is my plan. Once I fix the agent for the gate, I will dig into their code and try to figure out how to fix it. Please don't let that slow you down on what you need to get done in docs. | 22:28 |
| johnsom | I think the priority is still valid here: https://bugs.launchpad.net/octavia/+bugs?field.tag=target-mitaka | 22:29 |
| sbalukoff | johnsom: My basic impression after the e-mail thread yesterday on the docs and devs mailing list is that they don't have good guidelines on what should go in the manual and what shouldn't. In any case, it became very clear that housing docs in your own repo (like we do with Octavia) is totally fine-- but again, there are missing guidelines on what should go in the manual versus your own project doc tree. | 22:29 |
| johnsom | Yeah, especially since we straddle both neutron-lbaas and octavia. | 22:30 |
| sbalukoff | johnsom; Ok, my plan is to spend some time working on Octavia production installation docs. This might take several days, as we don't really have anything good right now (we just tell people to look at the devstack plugin, which is helpful but not exactly "polished") | 22:31 |
| johnsom | So, I guess I lean towards migrating out of docs into our repos over time. That will help with notices docs aren't done anyway. | 22:32 |
| sbalukoff | I've tried to make it really clear that we need guidance on which docs should go where. But I already have an idea of what we probably need in any case, so for now I'm just going to write it and shove it in the Octavia doc repo. We can hash out whether it should get into the main openstack manual some other time. But at least then we can claim that we have docs written. | 22:32 |
| *** ErnieP has quit IRC | 22:32 | |
| johnsom | Yep | 22:33 |
| johnsom | Ok, so are all of your non-unit test bugs fixed and awaiting review? | 22:33 |
| sbalukoff | We have a professional tech writer on staff now who has expressed eagerness about improving the openstack docs: And she was basically hit with the technical learning curve (cliff) on how to actually effectively work with the docs + unhelpful / clique-ish current contributors and has been pretty discouraged about improving the state of things there. | 22:34 |
| sbalukoff | I'll see if I can recruit her to help improve the docs in the Octavia project in the mean time. :) | 22:34 |
| johnsom | Yeah, go rst | 22:34 |
| johnsom | sbalukoff do you have a patch # for this? https://bugs.launchpad.net/octavia/+bug/1548212 | 22:35 |
| openstack | Launchpad bug 1548212 in octavia "Load balancer sits in PENDING_CREATE state much longer than necessary" [High,Fix committed] | 22:35 |
| sbalukoff | johnsom: All but the the URL validation using the standard library one... which didn't strike me as urgent. | 22:35 |
| johnsom | Yeah, I didn't tag that for Mitaka | 22:35 |
| sbalukoff | johnsom: Oh, that one was fixed by your REST timeout improvements. | 22:36 |
| johnsom | Oh, ok. I will close it out | 22:36 |
| *** piet has joined #openstack-lbaas | 22:42 | |
| dougwig | sbalukoff: i think if we structured our docs as devref like neutron docs/source/devref, and then pointed docs at that for content, we'd be totally fine. | 22:43 |
| dougwig | expcet api/cli, which is covered. | 22:43 |
| sbalukoff | dougwig: "we" as in Octavia, or "we" as in neutron-lbaas? | 22:43 |
| dougwig | i'd answer "yes", though i think n-l is in worse docs shape than octavia | 22:44 |
| sbalukoff | We don't really have much by way of neutron-lbaas docs right now. And what we do have was recently nuked from the manual (HA!) | 22:44 |
| sbalukoff | Yes, there's basically nothing in the n-l repo as far as docs. | 22:44 |
| sbalukoff | In Octavia we at least have API references, among other things. | 22:44 |
| johnsom | Either way, I advocate only doing docs in repo from now on. Too much effort was put in to play nice with docs to only have it deleted. | 22:46 |
| dougwig | did the lbaas api get deleted? | 22:50 |
| dougwig | no, it's still there. http://developer.openstack.org/api-ref-networking-v2-ext.html#lbaas-v2.0 | 22:51 |
| dougwig | good. | 22:51 |
| johnsom | No, it's still there. rotted I expect | 22:51 |
| sbalukoff | On the IBM side of things, doug-fish is also seeing if we can recruit a tech writer or two to help us improve the state of documentation in n-lbaas and octavia. | 22:51 |
| sbalukoff | So, that's great news! | 22:51 |
| johnsom | Excellent! Thank you | 22:52 |
| doug-fish | sbalukoff: wait - I thought we were on the same side now ?! | 22:52 |
| * doug-fish looks suspiciously at sbalukoff | 22:52 | |
| sbalukoff | doug-fish: Haha! | 22:53 |
| sbalukoff | johnsom: Yes, looking at the API ref, it doesn't have shared-pools or L7 stuff, and is subtly wrong in other ways. :/ | 22:55 |
| sbalukoff | Like... are members nested under pools in the n-l API? I thought all objects were root-level. | 22:56 |
| sbalukoff | Let me double-check that. | 22:56 |
| blogan | sbalukoff: they are under pools | 22:56 |
| johnsom | No, in V2 they are under | 22:56 |
| blogan | sbalukoff: and deleting a pool cascades it | 22:57 |
| sbalukoff | Oh! Huh. | 22:57 |
| blogan | you can blame or thank me for that | 22:57 |
| johnsom | blogan is the API path cooked? | 22:57 |
| sbalukoff | blogan: The more we make that API look like the Octavia API, the better, IMO. | 22:58 |
| blogan | johnsom: for? | 22:58 |
| blogan | johnsom: define cooked | 22:58 |
| johnsom | You are ready for the weekend huh? cascade delete | 22:58 |
| blogan | you northwesterner lingo | 22:58 |
| * johnsom looks guilty | 22:58 | |
| sbalukoff | Yeah, the lbaasv2 API reference doesn't look like it's been updated since Liberty. | 22:59 |
| sbalukoff | This is the "current" version of that document, right? | 22:59 |
| blogan | johnsom: the cascade delete, is still off /loadbalancer | 22:59 |
| blogan | i've voiced my opinion, but if yall want to veto it i need to know asap so i can work on the other way | 22:59 |
| johnsom | blogan I am asking if I should update the CLI code for whatever path was decided | 23:00 |
| blogan | lol we've got paralysis by indecision right now | 23:00 |
| blogan | dougwig: is the cascade_delete off the loadbalancer a strong NO? | 23:00 |
| johnsom | SoB, well, if we have the cores here we can start a vote. Otherwise I will defer the decision to dougwig our neutron-lbaas leader | 23:01 |
| blogan | dougwig: and is the create load balacner tree on the same /loadbalancers endpoint a strong nO? | 23:01 |
| dougwig | blogan: it's a soft no, because we can always alias that into a separate crud endpoint later. seems wrong, but eh. | 23:01 |
| blogan | if one of those is a strong NO then, that tells me go with new endpoint | 23:01 |
| dougwig | blogan: i'd find it really odd to overload create, after pondering it. | 23:02 |
| blogan | dougwig: does that odd feeling initiate a strong no? | 23:02 |
| dougwig | yes. | 23:02 |
| dougwig | but i'm one vote, not a dictator. | 23:02 |
| blogan | lt = dictator | 23:02 |
| johnsom | blogan +1 | 23:02 |
| johnsom | I was going to ask, since when? | 23:02 |
| dougwig | lt = janitor. | 23:03 |
| sbalukoff | Looks like the zuul gate is processing stuff again, but we may still have a problem with the py34 check. | 23:03 |
| sbalukoff | Probably need to wait to let the dust settle for a bit to know for sure. | 23:03 |
| blogan | dougwig: how do you feel about having a wonky extension just for cascade delete for M that doesn't expose any other methods, and then for Newton we add another extension to add all the missing methods (POST, PUT, GETS) | 23:03 |
| blogan | bc that makes me feel odd | 23:03 |
| dougwig | so lbtree-delete and lbtree? i wouldn't mind, tbh. | 23:04 |
| blogan | to have an endpoint /lb_tree/id that does not support GETs or PUTs and /lb_tree(s) that does not support POST | 23:04 |
| xgerman | what about lb-cactus? | 23:04 |
| blogan | dougwig: a /lbtree-delete endpoint? | 23:05 |
| blogan | please say no | 23:05 |
| dougwig | blogan: no, ext name. | 23:05 |
| dougwig | single endpoint between the two | 23:05 |
| blogan | oh okay | 23:05 |
| blogan | okay | 23:05 |
| blogan | the extension wiring is going ot be a mess | 23:05 |
| xgerman | I still don;t understand why we can’t have loadbalancer | 23:05 |
| dougwig | xgerman: because we already have it. | 23:06 |
| blogan | xgerman, dougwig: yall discuss that, i'll start working on the extension rewiring, we can always go back to the current PS | 23:06 |
| sbalukoff | xgerman: +1 Especially since these changes are backward-compatible. | 23:06 |
| sbalukoff | Er... at least on the create side. | 23:06 |
| xgerman | yep | 23:06 |
| xgerman | get-me-an-lb returns a load balancer — so it should be under loadbalancer | 23:07 |
| xgerman | sbalukoff we are not core here so just innocent bystanders when the murder REST | 23:07 |
| blogan | dougwig: one other thing, how is get-me-a-network doin git through the API? | 23:07 |
| xgerman | network-tree? | 23:07 |
| blogan | i saw we follow what they're doing | 23:07 |
| dougwig | ok, i feel like we have two sides of the fence based on tactical versus long-term here. forget mitaka. forget what exists. if we're making an API that operates on trees, what does it look like? i really do not care how it fits into the current api before i have the first answer. | 23:08 |
| dougwig | blogan: new endpoint | 23:08 |
| blogan | new endpoint it is | 23:08 |
| blogan | that sells it for me | 23:08 |
| xgerman | well, I guess I need to run for Neutron PTL so we can have nice things... | 23:09 |
| xgerman | probably the same thought process Donald Trump had | 23:09 |
| *** fnaval has joined #openstack-lbaas | 23:09 | |
| blogan | xgerman promises to build a wall at the Nova border and have Nova pay for it! | 23:10 |
| johnsom | Hahaa | 23:10 |
| dougwig | xgerman: you'd need a toupe | 23:10 |
| xgerman | I am already German that should so it ;-) | 23:10 |
| blogan | dougwig: they're calling get-me-a-network auto-allocated-topologies? | 23:11 |
| blogan | auto-allocated-lb it is | 23:11 |
| xgerman | lbaas/auto-allocated-lb | 23:12 |
| xgerman | ? | 23:12 |
| blogan | yep | 23:12 |
| xgerman | why can we do loadbalancer?type=autoallocated ? | 23:12 |
| blogan | we can still do lb-tree or whatever | 23:12 |
| xgerman | no, no trees — this is completely meaningless | 23:13 |
| xgerman | we return an lb so it should be called lb | 23:13 |
| xgerman | you don’t go to a bakery to get a car | 23:13 |
| blogan | we can return an lb_tree with a loadbalancer in it | 23:13 |
| blogan | in fact we are probably going to be forced into that | 23:13 |
| blogan | bc of neutron's extension magic | 23:13 |
| xgerman | arrgh | 23:13 |
| xgerman | I am glad I am not on the API team — I would need to medicate for my pain of doing unspeakable things to get resources | 23:14 |
| dougwig | xgerman: why do you feel so strongly that /loadbalancer should represent more than one type of object? that's where you lose me. | 23:14 |
| xgerman | get=good | 23:14 |
| blogan | dougwig: there's an argument to be made that its not one type of object, its the same | 23:15 |
| xgerman | yeah, don’t we return a loadbalancer | 23:15 |
| blogan | however, i will admit that if we called what is now loadbalancer, a vip, then we could call this a loadbalancer | 23:15 |
| xgerman | make a loadbalancer | 23:15 |
| xgerman | awesome!! | 23:15 |
| blogan | lol | 23:15 |
| blogan | salt in our wounds for the loadbalancer name | 23:15 |
| dougwig | xgerman: right, i wouldn't mind using loadbalancer in that context. but... *we already used that name for something else.* | 23:16 |
| blogan | damn you hindsight!!!!!!!! | 23:16 |
| xgerman | ok, so if you call get-me-an-lb without listener and members don’t you get the same lb as calling loadbalancer right now? | 23:16 |
| xgerman | So what we do right now is a special case | 23:17 |
| xgerman | of the bigger picture | 23:17 |
| dougwig | if you call a tree without a tree, i'd errorr it | 23:17 |
| blogan | what if its a partial tree? | 23:18 |
| sbalukoff | Woot! Looks like Leslie will be able to help us flesh out / correct documentation after next week (big deadlines for her until the 15th). So, we should get started in the mean time, but realize we've got actual published writers on the way to make it so our words good. :) | 23:18 |
| xgerman | what if all I want is a loadbalancer? | 23:18 |
| blogan | POST /loadbalancers | 23:18 |
| xgerman | but I want to use the fancy tree to do it | 23:18 |
| blogan | no reason to | 23:18 |
| blogan | if all you want is just the single loadbaalncer entity | 23:18 |
| xgerman | but I could? Or will it error if I don’t define listeners? | 23:19 |
| blogan | dougwig says it should | 23:19 |
| openstackgerrit | min wang proposed openstack/octavia: WIP----Replace the os.open method with safer way https://review.openstack.org/288208 | 23:20 |
| xgerman | that doesn’t make sense to me | 23:20 |
| xgerman | but so we are looking at lb_tree_id -> lb_id | 23:20 |
| dougwig | if you want just a loadbalancer use the loadbalancer API. if you insist on using the wrong mechanism for that, too bad. | 23:21 |
| blogan | dougwig: are you expecting the lb_tree to ahve its own ids? or accept loadbalancer ids? | 23:21 |
| *** piet has quit IRC | 23:21 | |
| sbalukoff | Just to throw another wrench into the clockwork: Technically it's not a tree, it's a graph. | 23:21 |
| sbalukoff | Because shared pools is a thing. | 23:22 |
| xgerman | dougwig I am trying to see if we are creating two endpoints for the same resource or if we artificially claim the other resource is different | 23:22 |
| blogan | lbgraph | 23:22 |
| blogan | done | 23:22 |
| xgerman | lbaas/auto-allocated-lb | 23:22 |
| xgerman | who knows it might morph into something else next | 23:22 |
| xgerman | ok, next question: if I create my stuff via the tree can I look at it with the current API or do I need to use the tree API | 23:27 |
| xgerman | is tree API LBaaS V3 without saying it? | 23:27 |
| dougwig | yes you can look at it, and yes it's v3 trying to shoehorn itself on top which is why we're arguing. | 23:29 |
| xgerman | so we want to create a parallel universe | 23:30 |
| blogan | versions? | 23:30 |
| xgerman | Fringe? | 23:31 |
| blogan | parallel universes = microversions | 23:31 |
| blogan | or really just an api running multiple versions of itself | 23:31 |
| xgerman | yep, that would be ideal... | 23:31 |
| dougwig | fuck parallel, let's go quantum. | 23:32 |
| blogan | and then go neutron? | 23:32 |
| dougwig | full circle. | 23:32 |
| xgerman | I always thought those get-me-* would be convenience functions to “orchestrate” getting something with a single call and I would then use the “normal” calls | 23:32 |
| xgerman | for everything else (and that single call might use the normal calls behind the scenes) | 23:33 |
| xgerman | never thought they would open up a parallel universe with anti-neutrinos | 23:33 |
| blogan | dont ever say the word orchestrate | 23:33 |
| xgerman | I put it in quotes | 23:34 |
| dougwig | hmm, if we're doing single create as a driver call, how do the objects get setup? single txn rollback? | 23:34 |
| xgerman | :-) | 23:34 |
| blogan | dougwig: when i create a port how many tables are inserted into? | 23:34 |
| *** kev0 has quit IRC | 23:34 | |
| blogan | same thing | 23:34 |
| blogan | and yes it would be a single txn rollback | 23:35 |
| blogan | well...if the driver call fails, the object is still in the db, but willbe marked as error | 23:36 |
| openstackgerrit | Merged openstack/octavia: Block deletion of pool in use by L7Policies https://review.openstack.org/285192 | 23:36 |
| blogan | lets ask the api working group | 23:38 |
| *** fawadkhaliq has quit IRC | 23:38 | |
| blogan | ah get-me-a-network already has a precedent | 23:38 |
| blogan | i'm gonig forward with this | 23:38 |
| blogan | with the coding of it | 23:39 |
| *** fawadkhaliq has joined #openstack-lbaas | 23:39 | |
| xgerman | also I haven’t seen a use case analysis nor a blueprint for that alternate universe — we only discussed that piecemeal until suddenly I am surrounded by graphs… sneaky... | 23:43 |
| blogan | too many metaphors for me to comprehend | 23:47 |
| openstackgerrit | Michael Johnson proposed openstack/neutron-lbaas: Fix the neutron-lbaas namespace driver gate https://review.openstack.org/288797 | 23:48 |
| johnsom | I doubt it is that easy, but it's a start | 23:48 |
| xgerman | nice | 23:48 |
| xgerman | blogan it took me days to figure out dougwig was pulling a fast one on me | 23:48 |
| blogan | well honestly, i should have realized the original implementation of cascade delete (the backwards incompatible one) was the wrong way | 23:49 |
| blogan | so blame me for that, in which case we'd have time to actually discuss this without being in a rush to figure it out | 23:50 |
| dougwig | johnsom: this gets it to the point where tempest runs: https://review.openstack.org/#/c/286332/ | 23:50 |
| johnsom | dougwig, well, I just did the same thing you did | 23:50 |
| johnsom | dougwig Should I take over that patch? | 23:51 |
| dougwig | johnsom: sure | 23:51 |
| xgerman | well, we had a fine solution in /cascade_delete which gets steamrolled into some tree/graph/whatever thing to open the door for a shoehorned V3 | 23:51 |
| dougwig | or yours, either way is fine. | 23:51 |
| blogan | xgerman: honestly, the right thing to do woudl be to defer all of this to N including cascade to delete, but we're trying to get it in | 23:53 |
| *** manishg has quit IRC | 23:53 | |
| blogan | how often do you use the API directly? if there's a cli there or a UI, you'd use that | 23:53 |
| johnsom | Yeah, not having this causes issues with the panels | 23:53 |
| *** Bjoern has quit IRC | 23:53 | |
| blogan | i get it | 23:53 |
| blogan | but i'm saying, if we didn't have that I'd totally just say it should be deferred | 23:53 |
| *** ducttape_ has quit IRC | 23:55 | |
| xgerman | well, we can’t do loadbalncer/<id>/delete_cascade to open the door for some yet undefined not understood graph project | 23:55 |
| blogan | hence, the right thing to do would be to defer both | 23:55 |
| xgerman | the right thing would be to do what makes sense now and then refactor in N | 23:56 |
| xgerman | but the waterfall overlords don’t allow us to do that | 23:56 |
| *** kevo has joined #openstack-lbaas | 23:57 | |
| sbalukoff | Heh! | 23:58 |
| sbalukoff | "cascade" "waterfall" | 23:58 |
| sbalukoff | I SEE WHAT YOU DID THERE! | 23:58 |
| xgerman | :-) | 23:58 |
Generated by irclog2html.py 2.14.0 by Marius Gedminas - find it at mg.pov.lt!