| *** elo has joined #akanda | 00:17 | |
| *** drwahl has quit IRC | 00:21 | |
| *** drwahl__ has joined #akanda | 00:21 | |
| openstackgerrit | Adam Gandelman proposed stackforge/akanda-rug: Various fixes to get drivers functional (WIP) https://review.openstack.org/223884 | 00:26 |
|---|---|---|
| adam_g | davidlenwell, ^ that might pass devstack now | 00:26 |
| *** elo has quit IRC | 00:33 | |
| *** qi has joined #akanda | 02:01 | |
| *** qi has quit IRC | 02:02 | |
| *** puranamr has joined #akanda | 02:17 | |
| davidlenwell | seems like its close.. thanks adam_g! | 02:39 |
| *** puranamr has quit IRC | 02:50 | |
| *** puranamr has joined #akanda | 03:11 | |
| *** puranamr has quit IRC | 03:15 | |
| *** puranamr has joined #akanda | 03:34 | |
| *** puranamr has quit IRC | 03:51 | |
| openstackgerrit | Adam Gandelman proposed stackforge/akanda-rug: Various fixes to get drivers functional (WIP) https://review.openstack.org/223884 | 05:23 |
| openstackgerrit | Adam Gandelman proposed stackforge/akanda-rug: Debug https://review.openstack.org/223932 | 06:17 |
| *** adam_g has quit IRC | 12:44 | |
| *** sarob has quit IRC | 12:48 | |
| *** ryanpetrello has quit IRC | 12:49 | |
| *** adam_g has joined #akanda | 12:50 | |
| *** adam_g has quit IRC | 12:51 | |
| *** adam_g has joined #akanda | 12:51 | |
| *** sarob has joined #akanda | 12:54 | |
| *** ryanpetrello has joined #akanda | 12:56 | |
| *** sarob has quit IRC | 13:07 | |
| *** ryanpetrello has quit IRC | 13:08 | |
| *** ryanpetrello has joined #akanda | 13:16 | |
| *** sarob has joined #akanda | 13:16 | |
| *** openstackgerrit has quit IRC | 14:31 | |
| *** openstackgerrit has joined #akanda | 14:31 | |
| *** openstackgerrit has quit IRC | 15:01 | |
| *** openstackgerrit has joined #akanda | 15:01 | |
| *** elo has joined #akanda | 16:05 | |
| *** elo has quit IRC | 16:15 | |
| *** elo has joined #akanda | 16:20 | |
| *** elo has quit IRC | 16:31 | |
| *** cleverdevil has joined #akanda | 17:13 | |
| adam_g | markmcclain, davidlenwell was trying to get some debug going on that patch last night at https://review.openstack.org/#/c/223932/ not sure whas up, its been working fine for me locally but i havent been starting with a fresh devststack, restacking now | 17:44 |
| davidlenwell | adam_g: I was looking at the console log from that patch just now. | 17:47 |
| adam_g | get_routers() seems to be returning nothing | 17:47 |
| markmcclain | yeah... was noticing that | 17:48 |
| davidlenwell | is it a context problem with the neutron client? | 17:49 |
| davidlenwell | might ad debug info to its state before calling that function | 17:49 |
| openstackgerrit | Adam Gandelman proposed stackforge/akanda-rug: Debug https://review.openstack.org/223932 | 17:50 |
| adam_g | davidlenwell, it shouldnt be, list_routers() gets called over rpc | 17:50 |
| davidlenwell | So that line you just added probably won't trigger .. because detailed defaults to true.. | 17:51 |
| adam_g | davidlenwell, router's pre_poopulate_hook calls /w detailed=False | 17:53 |
| davidlenwell | oh my bad .. I thought I left it blank | 17:54 |
| adam_g | hmm | 18:00 |
| adam_g | reproducing locally too | 18:00 |
| adam_g | oh | 18:04 |
| adam_g | i guess we dont ever even use pre-pop in the gate | 18:05 |
| adam_g | devstack starts ak-rug then creates network things as part of post-startup, so we rely on notifications to get going. i guess notifications are still not right in the driver patch | 18:05 |
| markmcclain | ah makes sense | 18:10 |
| * markmcclain wonders how to exercise pre-pop in the gate | 18:10 | |
| adam_g | markmcclain, well actually im a bit confused | 18:11 |
| adam_g | we create networks in both pre_start and post_start | 18:11 |
| markmcclain | don't we delete all of the routers? | 18:11 |
| adam_g | in pre start we scrub neutron | 18:11 |
| markmcclain | so the router notificaiton would not be seen until post_start | 18:11 |
| adam_g | and then re-create public net | 18:11 |
| adam_g | and mgt | 18:11 |
| adam_g | in post start we create 'thenet' | 18:11 |
| markmcclain | so the list should be empty | 18:12 |
| adam_g | markmcclain, im confused as to what we expect to be created on first start | 18:12 |
| adam_g | to exercise we could move creation stuff in post_start to pre_start, but then we lose exercising notifications. tho i have a patch up thats adds tests to stress that | 18:13 |
| adam_g | markmcclain, so the first/only router that ends up getting created is for thenet? | 18:13 |
| markmcclain | right it's for the net | 18:14 |
| adam_g | ok | 18:14 |
| adam_g | makes sense then | 18:14 |
| markmcclain | adam_g: so we would be able to test pre-pop | 18:14 |
| adam_g | ... if we create thenet as part of pre_pop | 18:14 |
| adam_g | derp, pre_start | 18:15 |
| markmcclain | sorry... forgot to send the 2nd part | 18:15 |
| markmcclain | we can test pre-pop in teh gate and test running 2nd rug | 18:15 |
| markmcclain | because if we create a 2nd rug after the first is running | 18:15 |
| markmcclain | pre-pop would get exercised too | 18:15 |
| markmcclain | *after the 1st is running and we've created any nets | 18:15 |
| markmcclain | +routers | 18:15 |
| adam_g | markmcclain, that could work | 18:16 |
| *** openstackgerrit has quit IRC | 18:16 | |
| adam_g | or we could just move it to pre_start to test in all cases, then theres a test case here that is entirely dependent on notifications https://review.openstack.org/#/c/219952/ | 18:16 |
| *** openstackgerrit has joined #akanda | 18:16 | |
| adam_g | okay this coffee shop is playing hootie and the blowfish. time to go home and look at notifications | 18:17 |
| markmcclain | adam_g: haha | 18:24 |
| adam_g | markmcclain, so im still confused--the router that we end up with has: | 18:29 |
| adam_g | | external_gateway_info | {"network_id": "d4146f68-e1ad-4efb-87d2-2b0195756ae9", "enable_snat": true, "external_fixed_ips": [{"subnet_id": "39911e81-9063-4cab-9de8-0ce61d36ef7c", "ip_address": "172.16.77.2"}, {"subnet_id": "fba27833-a6d4-427d-b67a-b277442c40a4", "ip_address": "fdee:9f85:83be::2"}]} | | 18:29 |
| adam_g | | ha | False | 18:29 |
| adam_g | those networks and subnets are the public network | 18:29 |
| markmcclain | right.. nothing attached to the private net | 18:30 |
| adam_g | markmcclain, so would that not be available for pre-pop, if public network is created pre-start of ak-rug? | 18:30 |
| markmcclain | actually that should be available | 18:32 |
| markmcclain | because we're not filtering out routers that are missing internal nets | 18:32 |
| davidlenwell | adam .. I believe I pointed out to you (and left in the commit message) that notifications was still wonky because of hte message format | 18:37 |
| adam_g | davidlenwell, https://review.openstack.org/#/c/223884/ attempts to finish that up | 18:40 |
| adam_g | markmcclain, i wonder if we're filtering them out by accident or via some side effect. it doesnt look like akanda reacts to anything until the demo thenet comes up with an interface on the public net | 18:44 |
| markmcclain | yeah... was looking to see if we're filtering too | 18:44 |
| markmcclain | but from the code we're calling list_routers which does not filter on teh neutron side | 18:45 |
| markmcclain | and the current driver code does not filter either | 18:46 |
| adam_g | markmcclain, so i still dont fully grok neutron's internal workflows, but it looks like when a tenant creates a net + subnet, it'll automatically get a router created for it | 18:47 |
| adam_g | whereas creating the public net alone wont do that | 18:47 |
| markmcclain | so the automated created is an old convenience hook we created for akanda deployments | 18:48 |
| markmcclain | basically to address the issue raised in the big thread in neutron right now | 18:48 |
| adam_g | markmcclain, is that hook in neutron-proper or akanda-neutron? | 18:50 |
| markmcclain | akanda-neutron | 18:53 |
| adam_g | so in a world where a single rug is managing multiple drivers, we want multiple drivers to be able to act on a single notification , right? | 19:08 |
| davidlenwell | hmm | 19:22 |
| davidlenwell | I don't know about that | 19:22 |
| adam_g | davidlenwell, would a port update/delete/create event not be relevant to both a router and a lbaas? | 19:42 |
| davidlenwell | hmm .. maybe.. depends on how fancy we want to get with lbaas | 19:44 |
| davidlenwell | would come in handy if we are doing service chaining | 19:45 |
| openstackgerrit | Adam Gandelman proposed stackforge/akanda-rug: Various fixes to get drivers functional (WIP) https://review.openstack.org/223884 | 20:01 |
| *** puranamr has joined #akanda | 20:42 | |
| *** puranamr has quit IRC | 20:47 | |
| *** puranamr has joined #akanda | 21:16 | |
| *** puranamr has quit IRC | 21:28 | |
| adam_g | markmcclain, still around? | 22:06 |
| markmcclain | adam_g: yes what's up | 22:07 |
| adam_g | markmcclain, thers nothing else that would be updating router state other than the synchronize_router_status() wrapper thing, right? | 22:08 |
| adam_g | oh nvm i wasn't loading code i was actually looking at | 22:13 |
| markmcclain | sorry had someone stop by... glad you figured out the issue | 22:23 |
| *** elo has joined #akanda | 23:27 | |
| *** elo has quit IRC | 23:47 | |
| *** elo has joined #akanda | 23:51 | |
| openstackgerrit | Adam Gandelman proposed stackforge/akanda-rug: Various fixes to get drivers functional (WIP) https://review.openstack.org/223884 | 23:56 |
| adam_g | closer | 23:56 |
| adam_g | markmcclain, davidlenwell | 23:56 |
| *** cleverdevil has quit IRC | 23:56 | |
Generated by irclog2html.py 2.14.0 by Marius Gedminas - find it at mg.pov.lt!