| johnsom | I am so curious why you need to add to the command line... | 00:01 |
|---|---|---|
| *** ducttape_ has joined #openstack-lbaas | 00:01 | |
| rm_work | net.ifnames=0 biosdevnames=0 | 00:04 |
| rm_work | gotta do it :/ | 00:04 |
| rm_work | also didn't we do away with the ubuntu specific elements? | 00:04 |
| rm_work | i guess not? haproxy-octavia / haproxy-octavia-ubuntu | 00:04 |
| johnsom | Not all of them yet. | 00:04 |
| rm_work | seems like they exist for everything still | 00:04 |
| johnsom | The directories are there for some just for backward compatibility (though I would probably just nuke them myself. | 00:05 |
| johnsom | Most are just elements-deps | 00:05 |
| johnsom | You can't just find the interface you need by mac or something? That is how we do it upstream. As it stands now, the amp doesn't care what the interfaces are named | 00:07 |
| rm_work | it's cloud-init | 00:08 |
| rm_work | if cloud-init doesn't see an eth0, it freaks out | 00:08 |
| rm_work | i think it might be something with this cloud | 00:08 |
| rm_work | but even stuff booted with octavia... | 00:08 |
| johnsom | Yeah, might be worth investigating a bit more (happy to look at console logs) as cloud-init doesn't get a eth0 on ubuntu 16.x, it's ens* The kernel renames it before cloud-init starts | 00:10 |
| johnsom | If you really want to get into cloud-init's brains, you can also mount the config drive: | 00:12 |
| johnsom | blkid -t LABEL="config-2" -odevice | 00:12 |
| johnsom | will give you the device | 00:12 |
| johnsom | just mount it like a cdrom. | 00:12 |
| rm_work | right, this cloud is non-ubuntu | 00:13 |
| johnsom | However, I will warn you, it can take a while to fish through all the stuff that ends up in there | 00:13 |
| rm_work | and they disable the new-naming on all their images | 00:13 |
| johnsom | So it should still be an eth*. | 00:14 |
| rm_work | yeah | 00:14 |
| rm_work | which is why i think their cloud-init is set up that way | 00:14 |
| johnsom | Yeah, I'm just saying, I don't think cloud-init should care | 00:14 |
| rm_work | i think they force their own cloud-init on every VM | 00:14 |
| johnsom | Oh | 00:15 |
| johnsom | Like Rackspace does | 00:15 |
| johnsom | bummer | 00:15 |
| rm_work | heh yeah | 00:15 |
| rm_work | i'll figure out a way around this | 00:15 |
| rm_work | i may have to write my own non-netns driver | 00:15 |
| rm_work | and amp-agent | 00:15 |
| rm_work | maybe can just rip a lot of the code from the liberty versions | 00:16 |
| rm_work | there's no custom networks here | 00:16 |
| rm_work | just one main one | 00:16 |
| rm_work | :( | 00:16 |
| johnsom | Are they maybe not supporting config drive? Maybe they are an EC2 shop only? | 00:16 |
| johnsom | I can't remember if I set it up to only accept config drive configs or not... | 00:17 |
| rm_work | config-drive def works | 00:18 |
| rm_work | the agent config is in place and it starts up | 00:18 |
| johnsom | Oh, ok | 00:18 |
| rm_work | just for whatever reason, the cloud-init scripts that run DEMAND an eth0 | 00:19 |
| rm_work | or else no networking will come up | 00:19 |
| rm_work | brb | 00:19 |
| johnsom | Maybe it's just another one of those "we have an ancient version" issues like git | 00:20 |
| *** amoralej|off has quit IRC | 00:28 | |
| *** ajo has quit IRC | 00:29 | |
| *** jschwarz has quit IRC | 00:29 | |
| *** fnaval has joined #openstack-lbaas | 00:39 | |
| *** ducttape_ has quit IRC | 00:40 | |
| *** diltram_ has joined #openstack-lbaas | 00:41 | |
| *** diltram has quit IRC | 00:45 | |
| *** fnaval_ has quit IRC | 00:45 | |
| *** reedip has quit IRC | 00:45 | |
| *** fnaval has quit IRC | 00:48 | |
| *** reedip has joined #openstack-lbaas | 00:55 | |
| *** ajo has joined #openstack-lbaas | 00:56 | |
| johnsom | Well, tomorrow I get to decide which code at fault for the quota scenario test failure. It think the quota code is getting run without a project ID... I.e. chicken/egg problem | 00:56 |
| *** ankur-gupta-f1 has quit IRC | 00:57 | |
| *** xgerman has quit IRC | 00:57 | |
| *** amoralej has joined #openstack-lbaas | 01:01 | |
| *** ankur-gupta-f1 has joined #openstack-lbaas | 01:07 | |
| *** xgerman has joined #openstack-lbaas | 01:07 | |
| openstackgerrit | Michael Johnson proposed openstack/octavia: Add quota support to Octavia https://review.openstack.org/360794 | 01:09 |
| *** amoralej has quit IRC | 01:19 | |
| *** yuanying has joined #openstack-lbaas | 01:20 | |
| *** ajo has quit IRC | 01:20 | |
| *** ducttape_ has joined #openstack-lbaas | 01:27 | |
| *** ajo has joined #openstack-lbaas | 01:32 | |
| *** jschwarz has joined #openstack-lbaas | 01:32 | |
| *** diogogmt has joined #openstack-lbaas | 01:35 | |
| *** amoralej has joined #openstack-lbaas | 01:38 | |
| *** catintheroof has quit IRC | 01:42 | |
| *** bana_k has quit IRC | 02:43 | |
| *** fnaval has joined #openstack-lbaas | 02:47 | |
| *** yuanying has quit IRC | 02:49 | |
| *** yuanying has joined #openstack-lbaas | 02:50 | |
| *** yuanying has quit IRC | 02:54 | |
| *** ducttape_ has quit IRC | 03:05 | |
| *** ducttape_ has joined #openstack-lbaas | 03:19 | |
| *** ducttape_ has quit IRC | 03:29 | |
| *** amotoki has joined #openstack-lbaas | 03:36 | |
| *** links has joined #openstack-lbaas | 03:42 | |
| *** yamamoto_ has joined #openstack-lbaas | 03:46 | |
| *** yuanying has joined #openstack-lbaas | 03:50 | |
| *** yuanying has quit IRC | 03:50 | |
| *** yuanying_ has joined #openstack-lbaas | 03:50 | |
| *** diogogmt has quit IRC | 03:54 | |
| *** fnaval has quit IRC | 04:03 | |
| *** ducttape_ has joined #openstack-lbaas | 04:29 | |
| *** ducttape_ has quit IRC | 04:34 | |
| *** yamamoto_ has quit IRC | 04:36 | |
| *** yamamoto_ has joined #openstack-lbaas | 04:43 | |
| *** bana_k has joined #openstack-lbaas | 04:52 | |
| *** yamamoto_ has quit IRC | 04:53 | |
| *** yamamoto_ has joined #openstack-lbaas | 04:57 | |
| *** yamamoto_ has quit IRC | 04:57 | |
| *** bana_k has quit IRC | 05:30 | |
| *** ducttape_ has joined #openstack-lbaas | 05:31 | |
| *** ducttape_ has quit IRC | 05:35 | |
| *** yamamoto_ has joined #openstack-lbaas | 05:39 | |
| *** amotoki has quit IRC | 05:47 | |
| *** pcaruana has joined #openstack-lbaas | 06:17 | |
| *** bana_k has joined #openstack-lbaas | 06:23 | |
| *** anilvenkata has joined #openstack-lbaas | 06:30 | |
| *** armax has joined #openstack-lbaas | 06:42 | |
| *** openstackgerrit has quit IRC | 06:48 | |
| *** yatin_away is now known as yatin | 06:54 | |
| *** tesseract has joined #openstack-lbaas | 06:58 | |
| *** tesseract has quit IRC | 06:58 | |
| *** tesseract has joined #openstack-lbaas | 06:59 | |
| *** tesseract is now known as Guest47337 | 06:59 | |
| *** amotoki has joined #openstack-lbaas | 07:08 | |
| *** jsheeren has joined #openstack-lbaas | 07:15 | |
| *** kobis has joined #openstack-lbaas | 07:16 | |
| *** openstackgerrit has joined #openstack-lbaas | 07:20 | |
| openstackgerrit | JingLiu proposed openstack/neutron-lbaas: Replace method vars().get to getattr https://review.openstack.org/413941 | 07:20 |
| *** armax has quit IRC | 07:24 | |
| *** amotoki_ has joined #openstack-lbaas | 07:26 | |
| *** amotoki has quit IRC | 07:28 | |
| *** ducttape_ has joined #openstack-lbaas | 07:32 | |
| *** ducttape_ has quit IRC | 07:37 | |
| *** ihrachys has joined #openstack-lbaas | 07:47 | |
| *** openstackgerrit has quit IRC | 07:48 | |
| *** ihrachys has quit IRC | 08:07 | |
| *** amotoki_ has quit IRC | 08:30 | |
| *** ducttape_ has joined #openstack-lbaas | 08:33 | |
| *** ducttape_ has quit IRC | 08:38 | |
| *** bana_k has quit IRC | 08:46 | |
| *** amotoki has joined #openstack-lbaas | 09:02 | |
| *** amotoki has quit IRC | 09:15 | |
| *** nmagnezi has joined #openstack-lbaas | 09:28 | |
| *** openstackgerrit has joined #openstack-lbaas | 09:39 | |
| openstackgerrit | Nir Magnezi proposed openstack/octavia: WIP - Fix the amphora-agent support for RH based Linux flavors https://review.openstack.org/331841 | 09:39 |
| *** amotoki has joined #openstack-lbaas | 09:51 | |
| *** amotoki has quit IRC | 10:20 | |
| *** gongysh has joined #openstack-lbaas | 10:39 | |
| *** ihrachys has joined #openstack-lbaas | 10:57 | |
| *** ducttape_ has joined #openstack-lbaas | 11:05 | |
| *** ducttape_ has quit IRC | 11:26 | |
| *** ducttape_ has joined #openstack-lbaas | 11:35 | |
| *** yamamoto_ has quit IRC | 11:46 | |
| *** ducttape_ has quit IRC | 11:52 | |
| *** gongysh has quit IRC | 12:08 | |
| *** anilvenkata has quit IRC | 12:17 | |
| *** yamamoto has joined #openstack-lbaas | 12:37 | |
| *** ducttape_ has joined #openstack-lbaas | 12:52 | |
| *** ducttape_ has quit IRC | 12:57 | |
| *** links has quit IRC | 13:18 | |
| *** chlong has joined #openstack-lbaas | 13:23 | |
| *** amoralej is now known as amoralej|lunch | 13:28 | |
| *** kobis has quit IRC | 13:41 | |
| *** armax has joined #openstack-lbaas | 14:03 | |
| *** kobis has joined #openstack-lbaas | 14:12 | |
| *** yatin has quit IRC | 14:16 | |
| *** ducttape_ has joined #openstack-lbaas | 14:19 | |
| *** kobis has quit IRC | 14:26 | |
| *** yatin has joined #openstack-lbaas | 14:29 | |
| *** kobis has joined #openstack-lbaas | 14:29 | |
| *** ducttape_ has quit IRC | 14:32 | |
| *** amoralej|lunch is now known as amoralej | 14:38 | |
| *** m-greene_ has joined #openstack-lbaas | 14:39 | |
| *** m-greene has quit IRC | 14:40 | |
| *** m-greene_ is now known as m-greene | 14:40 | |
| *** kobis has quit IRC | 14:45 | |
| *** nmagnezi has quit IRC | 14:51 | |
| *** ducttape_ has joined #openstack-lbaas | 14:56 | |
| *** m-greene has quit IRC | 15:08 | |
| *** jsheeren has quit IRC | 15:10 | |
| *** ducttape_ has quit IRC | 15:19 | |
| *** yamamoto has quit IRC | 15:25 | |
| *** ducttape_ has joined #openstack-lbaas | 15:26 | |
| xgerman | o/ | 15:29 |
| *** ducttape_ has quit IRC | 15:41 | |
| *** yamamoto has joined #openstack-lbaas | 15:48 | |
| *** yamamoto has quit IRC | 15:48 | |
| *** yamamoto has joined #openstack-lbaas | 15:48 | |
| *** madhu_ak has joined #openstack-lbaas | 15:54 | |
| *** Guest47337 has quit IRC | 16:03 | |
| johnsom | o/ | 16:04 |
| *** kobis has joined #openstack-lbaas | 16:07 | |
| openstackgerrit | Michael Johnson proposed openstack/octavia: Add quota support to Octavia https://review.openstack.org/360794 | 16:13 |
| *** ducttape_ has joined #openstack-lbaas | 16:15 | |
| openstackgerrit | Michael Johnson proposed openstack/octavia: Fix missing NovaServerGroupDelete https://review.openstack.org/409405 | 16:19 |
| *** yamamoto has quit IRC | 16:20 | |
| *** bana_k has joined #openstack-lbaas | 16:24 | |
| *** kobis has quit IRC | 16:28 | |
| *** ducttape_ has quit IRC | 16:34 | |
| *** kobis has joined #openstack-lbaas | 16:34 | |
| johnsom | xgerman FYI, I am also working on getting scenarios running local so I can clean up the scenario testing for quotas. | 16:49 |
| xgerman | cool | 16:49 |
| johnsom | So, if you have an ah-ha moments, let me know. | 16:49 |
| xgerman | I think the v2 scenarios ran but gave an error when accessing Octavia V2… | 16:49 |
| xgerman | nah, just tox -e | 16:50 |
| johnsom | Yeah, I'm up to the point where it only runs 2 tests | 16:50 |
| xgerman | mmh, how does it decide what to skip | 16:50 |
| xgerman | I hardcoded Octavia to be present in my tests | 16:51 |
| johnsom | Well, I think I need a tempest.conf | 16:51 |
| xgerman | https://review.openstack.org/#/c/413697/3/octavia/tests/tempest/config.py | 16:51 |
| xgerman | you can add stuff programmatically + I ran that inside devstack | 16:51 |
| *** fnaval has joined #openstack-lbaas | 16:52 | |
| *** kobis has quit IRC | 16:52 | |
| *** fnaval has quit IRC | 16:52 | |
| *** nmagnezi_ has joined #openstack-lbaas | 16:54 | |
| *** fnaval has joined #openstack-lbaas | 16:55 | |
| *** fnaval has quit IRC | 16:55 | |
| *** fnaval has joined #openstack-lbaas | 16:57 | |
| *** fnaval_ has joined #openstack-lbaas | 16:59 | |
| *** fnaval__ has joined #openstack-lbaas | 17:00 | |
| *** fnaval has quit IRC | 17:02 | |
| *** fnaval has joined #openstack-lbaas | 17:02 | |
| *** fnaval_ has quit IRC | 17:04 | |
| *** fnaval__ has quit IRC | 17:05 | |
| *** ducttape_ has joined #openstack-lbaas | 17:05 | |
| *** m-greene has joined #openstack-lbaas | 17:07 | |
| *** ducttape_ has quit IRC | 17:18 | |
| *** yamamoto has joined #openstack-lbaas | 17:20 | |
| *** mugsie_ has quit IRC | 17:22 | |
| *** yamamoto has quit IRC | 17:30 | |
| *** mugsie_ has joined #openstack-lbaas | 17:49 | |
| *** ihrachys has quit IRC | 17:56 | |
| *** kobis has joined #openstack-lbaas | 18:02 | |
| *** m-greene has quit IRC | 18:04 | |
| *** m-greene has joined #openstack-lbaas | 18:09 | |
| *** ducttape_ has joined #openstack-lbaas | 18:18 | |
| *** ducttape_ has quit IRC | 18:23 | |
| *** Matias is now known as Urahara | 18:24 | |
| *** kobis has quit IRC | 18:37 | |
| *** kobis has joined #openstack-lbaas | 18:45 | |
| *** kobis has quit IRC | 18:52 | |
| *** kobis has joined #openstack-lbaas | 18:55 | |
| *** amoralej is now known as amoralej|off | 19:01 | |
| openstackgerrit | OpenStack Proposal Bot proposed openstack/neutron-lbaas: Updated from global requirements https://review.openstack.org/414298 | 19:05 |
| *** ducttape_ has joined #openstack-lbaas | 19:06 | |
| openstackgerrit | OpenStack Proposal Bot proposed openstack/octavia: Updated from global requirements https://review.openstack.org/414272 | 19:06 |
| *** kobis has quit IRC | 19:07 | |
| *** ducttape_ has quit IRC | 19:23 | |
| *** nmagnezi_ has quit IRC | 19:30 | |
| *** yatin has quit IRC | 19:38 | |
| *** rcernin has joined #openstack-lbaas | 19:38 | |
| *** yatin has joined #openstack-lbaas | 19:50 | |
| johnsom | xgerman FYI, that global requirements has a major version jump in it, so not what I would call low-impact.... | 19:52 |
| xgerman | oh, taskflow | 19:53 |
| xgerman | need to look more closely next time | 19:53 |
| rm_work | ugh maybe the octavia scenarios would be useful for me ... | 20:17 |
| rm_work | what problems were you having johnsom? | 20:17 |
| *** ducttape_ has joined #openstack-lbaas | 20:24 | |
| openstackgerrit | German Eichberger proposed openstack/octavia: Add tempesttest for the N-LBaaS API https://review.openstack.org/413697 | 20:27 |
| xgerman | ^^ hours of pep8 appeasing… | 20:27 |
| rm_work | lol | 20:28 |
| rm_work | wonder if n-lbaas disabled some hacking rules that we didn't | 20:29 |
| xgerman | very likely | 20:29 |
| *** ducttape_ has quit IRC | 20:29 | |
| rm_work | wait what even IS a Nova Server Group? | 20:30 |
| rm_work | we can delete it before we delete the VM?? | 20:30 |
| xgerman | yep | 20:31 |
| xgerman | it controls anti-affinity | 20:31 |
| johnsom | rm_work I am working on the quotas patch. Mostly trying to figure out the right was to test under noauth since we need project-id for quotas. Right now we have hacked it into the requests, but that isn't the right answer | 20:34 |
| rm_work | hmm, dunno | 20:38 |
| rm_work | i mean | 20:38 |
| rm_work | if that works... :/ | 20:39 |
| xgerman | if we work under noauth we can’t really enforce quotas. Wouldn’t hose be enforced by n-lbaas instead? | 20:41 |
| johnsom | We are nlbaas now.. that is the point. I just think setting up a noauth "project" would make our tests better for both quota and policy | 20:45 |
| johnsom | Do it once and not on each test | 20:45 |
| xgerman | we couldn’t find another letter - n always reminds me of neutron | 20:47 |
| johnsom | rm_work BTW, on your question about NovaServerGroupDelete, this was just missing from the cascade delete flow where it is present in the main delete flow. Probably a refactor in order on the cascade delete flow so we don't have this kind of issue, but low priority.... | 20:51 |
| johnsom | xgerman I was not coining a new term "nlbaas", it still means neutron-lbaas to me, but what I was trying to say is with the merge we have to deal with our own quotas going forward. | 20:52 |
| xgerman | yep | 20:52 |
| xgerman | we need to enforce our own quotas… and if they come though neutrpn lbaas thsi will handle it | 20:53 |
| johnsom | Looking at other projects they all seem to handle this differently. One going as far as to create a noauth/test keystone middleware replacement for testing. | 20:53 |
| johnsom | Well, I expect for the pass through proxy with neutron lbaas we would remove the quota check for our calls in neutron. We don't want to be worried with two databases with quota info | 20:55 |
| xgerman | yep | 20:55 |
| xgerman | isn’t quota a neutron function and not lbaas?> | 20:56 |
| johnsom | I just hit the problem that now more scenario tests are running quota check code, so we really need to get the project_id stuff figured out in a reasonable way for the tests. | 20:56 |
| johnsom | Yes, that is why I said we would remove them in neutron. | 20:56 |
| johnsom | I.e pass them through as if the objects had no quota. | 20:57 |
| johnsom | The fun part is going to be the quota setting calls in neutron. Probably some custom code for that part. | 20:57 |
| xgerman | yeah, also they will have policies is neutron | 20:58 |
| xgerman | ‘so I would just make the internal endpoint quota free and the external one quota | 20:58 |
| xgerman | also did we model if somebody creates LB’s with Neutron LBaaS and then goes to Octavia with the new API… | 20:59 |
| xgerman | not that he gets twice the loadbalancers that way | 21:00 |
| johnsom | Since neutron-lbaas just passes the calls straight through, it's all one quota and all one code, so works the same | 21:00 |
| xgerman | so if somebody sets an LBaaS quota in neutron we carry that over? and if they are different last-one wins? | 21:01 |
| xgerman | or we just ignore neutron and make people configure octavia | 21:01 |
| johnsom | There is no octavia driver when using the pass through proxy | 21:01 |
| johnsom | pass through proxy is API request to neutron gets passed straight through to octavia API. | 21:02 |
| xgerman | yep, but quotas are configured in neutron.conf? | 21:02 |
| xgerman | so we ignore those? | 21:03 |
| xgerman | and the admin needs to establish them in Ocatvia | 21:03 |
| xgerman | same with policy files | 21:03 |
| johnsom | Yeah, in the future the quota for LBaaS objects in neutron-conf would be -1 | 21:03 |
| xgerman | ok, got it | 21:04 |
| rm_work | hmm so if i go through neutron-lbaas it creates a VIP for me... going direct to octavia, it seems we don't do that? vip is *required*? | 21:08 |
| johnsom | If one isn't passed in, it should create it | 21:09 |
| johnsom | rm_work Are you making a single call create? | 21:10 |
| rm_work | no | 21:10 |
| rm_work | also, VIP is definitely mandatory :/ | 21:11 |
| johnsom | Ah, I was confusing. I was thinking port | 21:13 |
| johnsom | http://docs.openstack.org/developer/octavia/api/octaviaapi.html#create-load-balancer | 21:13 |
| johnsom | Going through neutron-lbaas, but default, creates the VIP port for us and passes in a port ID for the vip. | 21:14 |
| xgerman | I thought we can create our own VIP in Octavia… blogan worked on some flows | 21:15 |
| johnsom | Yeah, he worked on it to work as I described above. Before we always created the VIP ports ourselves | 21:15 |
| xgerman | but we still do that if none is submitted? | 21:16 |
| johnsom | Right if the v1 api just gets a subnet, it will create the port | 21:16 |
| johnsom | https://github.com/openstack/neutron-lbaas/blob/master/neutron_lbaas/drivers/octavia/driver.py#L54 | 21:17 |
| rm_work | subnetID is ... | 21:18 |
| rm_work | the subnet of the FIP's IP? | 21:18 |
| johnsom | Yeah, subnet ID you want your VIP on | 21:18 |
| xgerman | but not FIP | 21:18 |
| xgerman | FIP is on the router forwarding to VIP | 21:18 |
| xgerman | aka in some net which is connected to the router | 21:19 |
| johnsom | Yeah, none of this manages FIPs. (FIP== the nat IP) | 21:19 |
| blogan | xgerman: johnsom answered correctly | 21:21 |
| xgerman | I would have expected nothing less from him ;-) | 21:21 |
| johnsom | Ha! | 21:23 |
| johnsom | I just had to go deep on that stuff to fix the failover stuff | 21:24 |
| *** ducttape_ has joined #openstack-lbaas | 21:25 | |
| *** ducttape_ has quit IRC | 21:30 | |
| rm_work | urg wait so | 21:34 |
| rm_work | I create a FIP | 21:34 |
| rm_work | it's on the FIP network | 21:34 |
| rm_work | it has an IP | 21:34 |
| xgerman | and then you associate that FIP with the VIP | 21:35 |
| rm_work | wait | 21:35 |
| rm_work | what is a VIP | 21:35 |
| rm_work | I thought the FIP was the VIP | 21:35 |
| rm_work | hold on | 21:35 |
| xgerman | well, we spin the LB up in the mgmt net and the user’s subnet | 21:36 |
| rm_work | uhh so | 21:36 |
| xgerman | it will need a VIP in the user’s subnet | 21:36 |
| rm_work | to create a LB | 21:36 |
| xgerman | and then if the subnet isn’t connected to the Internet | 21:36 |
| rm_work | https://gist.github.com/anonymous/f59d96918f3103ad5651 | 21:36 |
| xgerman | you need some FIP on the ext-net which get associated with the VIP | 21:36 |
| rm_work | i'm following that (stephen's thing) | 21:36 |
| rm_work | first command... i need to pass in an IP and a subnet | 21:37 |
| rm_work | what are those from? | 21:37 |
| xgerman | thi is the VIP | 21:37 |
| rm_work | ok, what type of object is the VIP? | 21:37 |
| xgerman | it’s just an IP | 21:37 |
| rm_work | ... | 21:37 |
| rm_work | but I don't have an IP anywhere yet | 21:37 |
| xgerman | well, he is passing on IP, port and net | 21:38 |
| rm_work | what command do I run to create an IP? | 21:38 |
| rm_work | what type of openstack object is it | 21:38 |
| xgerman | so I think if you only pass in subnet we create a port | 21:38 |
| rm_work | i can't just pick an IP address out of thin air | 21:38 |
| rm_work | it has to be assigned to me somehow | 21:38 |
| rm_work | is it a port? | 21:39 |
| rm_work | do I just need to create a neutron port on some subnet? | 21:39 |
| xgerman | that might be sufficeint - neutron assigns ips often via dhcp | 21:40 |
| xgerman | but I always had Octavia deal with that ;-) | 21:40 |
| rm_work | yeah but ... i can't, right? | 21:40 |
| rm_work | this is mandatory to create the highest level object | 21:40 |
| rm_work | I have to pass in at LEAST "ip_address" and "subnet" for VIP | 21:40 |
| xgerman | subnet should be all you need | 21:40 |
| rm_work | it requires more | 21:42 |
| rm_work | if i just provide subnet it flips out, AFAICT | 21:42 |
| rm_work | sec | 21:42 |
| rm_work | ok yeah so creating a port and passing that info seems right | 21:45 |
| xgerman | https://github.com/openstack/octavia/blob/master/octavia/network/drivers/neutron/allowed_address_pairs.py#L324-L353 | 21:45 |
| xgerman | we have code to create a port AFAIK | 21:45 |
| rm_work | then ... | 21:47 |
| rm_work | why is VIP a required attribute on create of a LB | 21:47 |
| xgerman | well, that code won’t work unless you pass aport in… so umh | 21:48 |
| xgerman | likely needs rafactor | 21:49 |
| rm_work | yeah this seems un-like the neutron API :P | 21:49 |
| johnsom | Hang on here. | 21:49 |
| johnsom | So, you need either a port or a subnet. IP is (should be) optional. | 21:49 |
| rm_work | when i didn't provide one, it got an exception and bailed | 21:49 |
| rm_work | i can try to reproduce really quick | 21:50 |
| johnsom | If you pass in a port_id, we use it for the VIP. If not we create one for you | 21:50 |
| xgerman | johnsom you need both | 21:50 |
| xgerman | https://www.irccloud.com/pastebin/OrRJXCZW/ | 21:50 |
| xgerman | there seems to be a bug in the first of | 21:50 |
| johnsom | That is odd, we shouldn't need the IP. | 21:50 |
| xgerman | we require a port-id but the message says it shoudl be eitehr - or | 21:50 |
| xgerman | and the code looks like it, too | 21:50 |
| rm_work | ah that's right | 21:51 |
| rm_work | no it is | 21:51 |
| rm_work | it's not A and not B | 21:51 |
| rm_work | meaning, one has to exist | 21:51 |
| johnsom | That logic is correct | 21:51 |
| rm_work | i read it wrong first but second reading it made sense | 21:51 |
| johnsom | Right, port_id or subnet_id are required | 21:51 |
| xgerman | oh, ok | 21:51 |
| xgerman | then all you would need is a subnet | 21:51 |
| johnsom | But I don't think we require an IP specified. We can allocate that from the subnet if one isn't given, so if that fails it is a bug | 21:52 |
| xgerman | we shouldn;’t even require a port | 21:52 |
| johnsom | We don't require a port | 21:52 |
| xgerman | rm_work says we do | 21:52 |
| johnsom | We require a port_id OR a subnet. | 21:52 |
| rm_work | oh hmm... one sec, might have worked that time without the ip | 21:52 |
| rm_work | i meant it was requiring subnet and IP | 21:53 |
| rm_work | but | 21:53 |
| rm_work | trying to repro | 21:53 |
| xgerman | the way it should work is that you cna pass in a subnet and it should be happy | 21:53 |
| rm_work | yeah maybe it is working now... maybe something was wrong with the subnet i was trying befor | 21:55 |
| xgerman | don’t scare us like that :-)_ | 21:56 |
| johnsom | Ha, I was going to try it, but I hit my quota code... so, yeah. | 21:56 |
| xgerman | at least something works | 21:57 |
| rm_work | 2016-12-22 14:56:59.797 4238 ERROR oslo_messaging.rpc.server return data_models.Vip(ip_address=fixed_ip.ip_address, | 21:57 |
| rm_work | 2016-12-22 14:56:59.797 4238 ERROR oslo_messaging.rpc.server AttributeError: 'NoneType' object has no attribute 'ip_address' | 21:57 |
| rm_work | i think that's maybe because the subnet could be out of IPs? | 21:57 |
| * rm_work guesses wildly | 21:57 | |
| rm_work | so it might just be a bad subnet | 21:57 |
| rm_work | yeah looks like. | 22:00 |
| rm_work | hmm, trying to use the management subnet as the vip subnet and it breaks network connectivity when it plugs the VIP, it seems | 22:01 |
| rm_work | :/ | 22:01 |
| rm_work | i know that's not ideal but i don't have access to multiple working subnets :/ | 22:01 |
| johnsom | Yeah, I just tested, it works without an IP | 22:02 |
| xgerman | Today’s reading: https://medium.com/vimeo-engineering-blog/improving-load-balancing-with-a-new-consistent-hashing-algorithm-9f1bd75709ed#.48ssreeiw | 22:03 |
| johnsom | Interesting. Oslo added a consistent hash library recently too. I think it is very basic though | 22:05 |
| xgerman | nice | 22:05 |
| johnsom | Yeah, 3-4 projects are all using some form of consistent hashing | 22:07 |
| xgerman | this is awesome. This algo is in haproxy 1.7 | 22:08 |
| xgerman | now we need to add that to the pool ;-) | 22:11 |
| rm_work | ugh ssl handshake failure between amp and controller WHY | 22:12 |
| *** Kiall has joined #openstack-lbaas | 22:13 | |
| johnsom | rm_work you need to sacrifice something to OpenStack | 22:14 |
| rm_work | ah because i'm dumb, that's why | 22:14 |
| rm_work | mixed config sources | 22:15 |
| xgerman | I guess he already sacrificed sanity :-) | 22:15 |
| rm_work | so close... | 22:15 |
| rm_work | OH SHIT | 22:17 |
| rm_work | it went active :3 | 22:17 |
| rm_work | WHAT HAS SCIENCE WRAUGHT?! | 22:17 |
| * rm_work is feeling super ridiculous after struggling with getting everything working for two weeks | 22:17 | |
| *** madhu_ak has quit IRC | 22:24 | |
| *** chlong has quit IRC | 22:34 | |
| *** chlong has joined #openstack-lbaas | 22:41 | |
| openstackgerrit | Michael Johnson proposed openstack/octavia: Add quota support to Octavia https://review.openstack.org/360794 | 22:53 |
| *** ducttape_ has joined #openstack-lbaas | 22:54 | |
| *** m-greene has quit IRC | 22:59 | |
| *** Urahara is now known as Matias | 23:15 | |
| rm_work | why is "weight" required on members? | 23:22 |
| rm_work | shouldn't it just default to "1" or something? | 23:22 |
| rm_work | it doesn't even make sense for ROUND_ROBIT | 23:22 |
| rm_work | *ROBIN | 23:22 |
| rm_work | oh wait, did we make ROUND_ROBIN and WEIGHTED_ROUND_ROBIN separate or is it just always weighted | 23:23 |
| *** ducttape_ has quit IRC | 23:25 | |
| johnsom | weight 0 means it doesn't get new connections | 23:29 |
| johnsom | weight is not required, it defaults to 1 | 23:30 |
| johnsom | FYI, the gates are broken to some degree. OSC client errors doing keystone things. | 23:31 |
| johnsom | A bunch of projects are impacted | 23:32 |
| rm_work | johnsom: http://docs.openstack.org/developer/octavia/api/octaviaapi.html#create-member according to that, weight is required | 23:37 |
| rm_work | is it just wrong in the docs? | 23:37 |
| johnsom | rm_work Yes, I think so: https://github.com/openstack/octavia/blob/master/octavia/api/v1/types/member.py#L40 | 23:38 |
| rm_work | yeah, seems so, was able to create one without specifying | 23:38 |
| rm_work | i thought the docs were generated | 23:38 |
| johnsom | Nope | 23:38 |
| rm_work | ah also subnet ISN'T required | 23:39 |
| rm_work | is that a deviation from n-lbaas? | 23:39 |
| johnsom | That is interesting. I but it pukes without it though | 23:39 |
| rm_work | hmm, no, n-lbaas api doc says subnet_id isn't required... wtf | 23:40 |
| rm_work | ahhh | 23:40 |
| rm_work | it's a client bug | 23:40 |
| rm_work | neutron lbaas-member-create requires subnet | 23:40 |
| rm_work | but the API doesn't actually | 23:40 |
| rm_work | that explains my confusion | 23:41 |
| johnsom | I think it should be required in the api | 23:41 |
| johnsom | Not sure, would have to dig through the code. | 23:42 |
| rm_work | i kept trying to USE it, finding it required... being confused... then later looking it up and thinking "i thought it WAS required, must have been mistaken" | 23:42 |
| rm_work | and then trying again and it's required in the client | 23:42 |
| rm_work | it shouldn't be required IMO | 23:42 |
| rm_work | if you know you can get out via the VIP subnet | 23:42 |
| rm_work | you shouldn't need to specify one for the member | 23:42 |
| rm_work | especially given the way this is set up in some clouds, it'd be very confusing | 23:43 |
| rm_work | as we have a bunch of subnets connected via routing, but not directly pluggable | 23:43 |
| rm_work | so like, my member might very well be on abcd-efg subnet, but i'd need to have the amp bind on subnet qwer-tyuip | 23:44 |
| rm_work | and there's no way i'd know that as an end user | 23:44 |
| rm_work | in fact, i'm probably going to have to figure out a way to have octavia select a default vip subnet, and make it not require a VIP object | 23:44 |
| rm_work | because the user won't have any way of knowing what networks they can get a VIP on | 23:45 |
| rm_work | everything is just routed together | 23:45 |
| rm_work | ugh, network setups always throw a wrench in everything because everyone does it totally differently in their clouds | 23:45 |
| rm_work | johnsom: unless, do we do anything to the routes that would prevent traffic flow from going out the VIP interface to reach a member? | 23:49 |
| johnsom | rm_work No, I don't think we do anymore | 23:51 |
| rm_work | ok... | 23:51 |
| rm_work | then subnet_id should absolutely be optional | 23:51 |
| johnsom | I get your point | 23:51 |
| rm_work | and I need to figure out how to make this VIP thing work | 23:52 |
| *** ducttape_ has joined #openstack-lbaas | 23:59 | |
Generated by irclog2html.py 2.14.0 by Marius Gedminas - find it at mg.pov.lt!