| *** threestrands has joined #openstack-lbaas | 00:03 | |
| *** threestrands has quit IRC | 00:03 | |
| *** threestrands has joined #openstack-lbaas | 00:03 | |
| *** shananigans has quit IRC | 00:40 | |
| *** shananigans has joined #openstack-lbaas | 00:40 | |
| *** longkb has joined #openstack-lbaas | 00:43 | |
| *** yamamoto has joined #openstack-lbaas | 00:52 | |
| *** threestrands_ has joined #openstack-lbaas | 01:18 | |
| *** threestrands_ has quit IRC | 01:18 | |
| *** threestrands_ has joined #openstack-lbaas | 01:18 | |
| *** threestrands_ has quit IRC | 01:19 | |
| *** threestrands_ has joined #openstack-lbaas | 01:20 | |
| *** hongbin has joined #openstack-lbaas | 01:20 | |
| *** threestrands_ has quit IRC | 01:21 | |
| *** phuoc has joined #openstack-lbaas | 01:21 | |
| *** threestrands_ has joined #openstack-lbaas | 01:21 | |
| *** threestrands_ has quit IRC | 01:21 | |
| *** threestrands_ has joined #openstack-lbaas | 01:21 | |
| *** threestrands has quit IRC | 01:21 | |
| *** phuoc_ has quit IRC | 01:24 | |
| *** phuoc_ has joined #openstack-lbaas | 01:52 | |
| *** sapd_ has quit IRC | 01:54 | |
| *** sapd has joined #openstack-lbaas | 01:55 | |
| *** phuoc has quit IRC | 01:55 | |
| *** phuoc has joined #openstack-lbaas | 02:02 | |
| *** phuoc_ has quit IRC | 02:04 | |
| *** phuoc_ has joined #openstack-lbaas | 02:16 | |
| *** phuoc has quit IRC | 02:18 | |
| *** sapd_ has joined #openstack-lbaas | 02:22 | |
| *** sapd has quit IRC | 02:22 | |
| *** annp has joined #openstack-lbaas | 02:30 | |
| *** phuoc has joined #openstack-lbaas | 02:40 | |
| *** phuoc_ has quit IRC | 02:44 | |
| *** phuoc_ has joined #openstack-lbaas | 02:46 | |
| *** phuoc has quit IRC | 02:49 | |
| *** phuoc has joined #openstack-lbaas | 02:50 | |
| *** phuoc_ has quit IRC | 02:52 | |
| lxkong | hi johnsom, does octaiva amphora image support fedora now? | 02:53 |
|---|---|---|
| lxkong | i met with a problem when i was building amphora image with `-i fedora` | 02:53 |
| lxkong | the error log: https://paste.ubuntu.com/p/mhKksYcQG6/ | 02:53 |
| *** phuoc has quit IRC | 03:08 | |
| *** sapd__ has joined #openstack-lbaas | 03:17 | |
| *** sapd_ has quit IRC | 03:18 | |
| *** ramishra has joined #openstack-lbaas | 03:20 | |
| *** gans has joined #openstack-lbaas | 03:25 | |
| *** gans has quit IRC | 03:27 | |
| *** atoth has quit IRC | 03:35 | |
| *** annp has quit IRC | 03:56 | |
| *** annp has joined #openstack-lbaas | 03:56 | |
| *** sapd has joined #openstack-lbaas | 04:03 | |
| *** hongbin has quit IRC | 04:27 | |
| *** phuoc has joined #openstack-lbaas | 04:34 | |
| *** phuoc has quit IRC | 04:40 | |
| *** links has joined #openstack-lbaas | 05:00 | |
| *** AlexStaf has quit IRC | 05:10 | |
| *** phuoc has joined #openstack-lbaas | 05:20 | |
| johnsom | lxkong: looks like a bug. I added that element to remove those packages recently somince | 05:34 |
| johnsom | Since ubuntu’s image size keeps growing. That is on stable branch though. | 05:35 |
| johnsom | We don’t gate on fedora, just centos on the redhat side | 05:35 |
| johnsom | It must be a missing pkg-map. | 05:36 |
| lxkong | johnsom: i am using queens stable | 05:46 |
| johnsom | Yeah, those package exceptions are only on the stable branches. I solved it a better way on master, just could not backport it due to policy. | 05:47 |
| johnsom | You could build a master branch image if you really need fedora | 05:49 |
| *** yboaron_ has joined #openstack-lbaas | 06:13 | |
| *** abaindur has quit IRC | 06:21 | |
| *** ispp has joined #openstack-lbaas | 06:24 | |
| *** longkb has quit IRC | 06:26 | |
| *** sapd has quit IRC | 06:26 | |
| *** longkb has joined #openstack-lbaas | 06:28 | |
| *** annp has quit IRC | 06:30 | |
| *** annp has joined #openstack-lbaas | 06:30 | |
| *** longkb has quit IRC | 06:37 | |
| *** longkb has joined #openstack-lbaas | 06:38 | |
| *** peereb has joined #openstack-lbaas | 07:05 | |
| *** tesseract has joined #openstack-lbaas | 07:06 | |
| *** velizarx has joined #openstack-lbaas | 07:07 | |
| *** rcernin has quit IRC | 07:08 | |
| *** nmanos has joined #openstack-lbaas | 07:20 | |
| *** tesseract has quit IRC | 07:25 | |
| *** tesseract has joined #openstack-lbaas | 07:27 | |
| *** kobis has joined #openstack-lbaas | 07:30 | |
| *** ispp has quit IRC | 07:32 | |
| nmagnezi | lxkong, If it's possible for you, I think you better favor centos over fedora amp (being tested in CI) | 07:35 |
| lxkong | nmagnezi: thanks for your advise. But for some reason(relating to openssl) we can only use fedora based image in our CI :-( | 07:36 |
| lxkong | johnsom: can you tell me the patch, maybe i can backport to our internal repo | 07:37 |
| *** velizarx has quit IRC | 07:37 | |
| *** nmanos has left #openstack-lbaas | 07:41 | |
| *** velizarx has joined #openstack-lbaas | 07:48 | |
| *** ispp has joined #openstack-lbaas | 07:54 | |
| *** rpittau has joined #openstack-lbaas | 07:57 | |
| *** zigo has quit IRC | 08:03 | |
| *** keithmnemonic[m] has quit IRC | 08:05 | |
| *** zigo has joined #openstack-lbaas | 08:05 | |
| *** threestrands_ has quit IRC | 08:12 | |
| *** keithmnemonic[m] has joined #openstack-lbaas | 08:27 | |
| *** yamamoto has quit IRC | 08:28 | |
| *** ramishra has quit IRC | 08:39 | |
| *** ispp has quit IRC | 08:44 | |
| *** ramishra has joined #openstack-lbaas | 08:45 | |
| lxkong | johnsom: is it this one https://review.openstack.org/#/c/559416/? | 08:49 |
| lxkong | johnsom: btw, can i use master version(or even Rocky) amphora to work with queens octavia? I suppose the communication with amphora is backward compatible, right? | 08:51 |
| *** yamamoto has joined #openstack-lbaas | 09:12 | |
| *** ramishra has quit IRC | 09:15 | |
| *** ramishra has joined #openstack-lbaas | 09:17 | |
| nmagnezi | lxkong, it *should* work, but I would advise against using master code in production (as opposed to stable versions), but on the other hand I know some folks are doing just that. | 09:20 |
| *** kobis has quit IRC | 09:21 | |
| *** kobis has joined #openstack-lbaas | 09:22 | |
| *** abaindur has joined #openstack-lbaas | 09:22 | |
| *** kobis has quit IRC | 09:26 | |
| *** ispp has joined #openstack-lbaas | 09:46 | |
| *** kobis has joined #openstack-lbaas | 09:50 | |
| openstackgerrit | Nguyen Van Trung proposed openstack/octavia master: Follow the new PTI for document build https://review.openstack.org/580983 | 10:04 |
| *** longkb has quit IRC | 10:10 | |
| *** rraja has joined #openstack-lbaas | 10:50 | |
| lxkong | nmagnezi: maybe I should add some context here. We are running Pike in production, and I am testing Queens, so i need fedora amphora image of queens for CI, and johnsom said there is some package issue for stable branch and it's impossible to backport the fix. What i'm thinking now is to either build fedora image of master branch or backport the fix to queens in our internal repo. | 11:29 |
| lxkong | it should be "it's impossible to backport the fix for upstream" | 11:30 |
| *** atoth has joined #openstack-lbaas | 11:39 | |
| *** yamamoto has quit IRC | 11:57 | |
| *** kobis has quit IRC | 12:03 | |
| openstackgerrit | ZhaoBo proposed openstack/octavia master: UDP jinja template https://review.openstack.org/525420 | 12:09 |
| openstackgerrit | ZhaoBo proposed openstack/octavia master: WIP:UDP for [2] https://review.openstack.org/529651 | 12:09 |
| openstackgerrit | ZhaoBo proposed openstack/octavia master: UDP for [3][5][6] https://review.openstack.org/539391 | 12:09 |
| *** kobis has joined #openstack-lbaas | 12:15 | |
| *** yamamoto has joined #openstack-lbaas | 12:16 | |
| *** yamamoto has quit IRC | 12:16 | |
| *** amuller has joined #openstack-lbaas | 12:17 | |
| *** yamamoto has joined #openstack-lbaas | 12:18 | |
| *** yboaron_ has quit IRC | 12:44 | |
| *** velizarx has quit IRC | 12:45 | |
| *** yboaron_ has joined #openstack-lbaas | 12:45 | |
| *** velizarx has joined #openstack-lbaas | 12:49 | |
| *** wolsen has quit IRC | 13:18 | |
| *** wolsen has joined #openstack-lbaas | 13:20 | |
| johnsom | lxkong: I think we can fix fedora on the stable branches, it just needs a pkg-map change. This will still be different than master fix, but should work. I will try to get a patch up today for you. | 13:29 |
| johnsom | We just need to exclude those two package removals from fedora. | 13:30 |
| *** fnaval has joined #openstack-lbaas | 13:44 | |
| *** links has quit IRC | 14:18 | |
| openstackgerrit | Allen proposed openstack/octavia master: [doc] Add the missing markup for the hyperlink title https://review.openstack.org/581046 | 14:26 |
| *** AlexStaf has joined #openstack-lbaas | 14:53 | |
| *** yamamoto has quit IRC | 15:01 | |
| *** peereb has quit IRC | 15:03 | |
| *** AlexStaf has quit IRC | 15:07 | |
| *** yamamoto has joined #openstack-lbaas | 15:17 | |
| *** yamamoto has quit IRC | 15:22 | |
| *** velizarx has quit IRC | 15:24 | |
| * xgerman_ reading scrollback | 15:27 | |
| *** rraja has quit IRC | 15:37 | |
| *** kobis has quit IRC | 15:42 | |
| *** ivve has joined #openstack-lbaas | 15:44 | |
| *** yamamoto has joined #openstack-lbaas | 15:46 | |
| *** yamamoto has quit IRC | 15:46 | |
| johnsom | lxkong https://review.openstack.org/581073 | 16:02 |
| johnsom | Or Pike: https://review.openstack.org/581074 | 16:03 |
| *** ispp has quit IRC | 16:13 | |
| *** ramishra has quit IRC | 16:15 | |
| openstackgerrit | Michael Johnson proposed openstack/octavia master: Implement provider drivers - Driver Library https://review.openstack.org/571358 | 16:18 |
| *** abaindur has quit IRC | 16:25 | |
| *** velizarx has joined #openstack-lbaas | 16:38 | |
| openstackgerrit | Michael Johnson proposed openstack/neutron-lbaas master: Neutron-LBaaS to Octavia migration tool https://review.openstack.org/578942 | 16:40 |
| openstackgerrit | Michael Johnson proposed openstack/neutron-lbaas master: Neutron-LBaaS to Octavia migration tool https://review.openstack.org/578942 | 16:42 |
| *** yamamoto has joined #openstack-lbaas | 16:47 | |
| openstackgerrit | Michael Johnson proposed openstack/neutron-lbaas master: Neutron-LBaaS to Octavia migration tool https://review.openstack.org/578942 | 16:49 |
| *** yamamoto has quit IRC | 16:57 | |
| jiteka | Hello, I ran into a problem today in my devstack lab, my first LB took longer than expected to spawn amphora VM probably due to lack of cached image as it was the first one | 17:01 |
| jiteka | As result, even if now my amphora VM is running, amphora status is still "BOOTING" and amphora VM is active and reachable via ping on his lb-mgmt-net interface | 17:01 |
| jiteka | So I have 2 question : | 17:01 |
| jiteka | 1. Is it possible to force a refresh of the state to resume where it failed (or got stuck) | 17:01 |
| jiteka | 2. What need to be cleaned on the octavia DB after deleting that stuck amphora VM ? (or is any other way exist to delete PENDING_CREATE loadbalancer ?) | 17:01 |
| *** velizarx has quit IRC | 17:12 | |
| *** tesseract has quit IRC | 17:15 | |
| *** sapd has joined #openstack-lbaas | 17:21 | |
| xgerman_ | jiteka: this seems odd - the system should move it to error after an hour or so — | 17:30 |
| *** sapd has quit IRC | 17:33 | |
| openstackgerrit | German Eichberger proposed openstack/octavia master: [WIP] Switch amphora agent to use privsep https://review.openstack.org/549295 | 17:41 |
| *** sapd has joined #openstack-lbaas | 17:41 | |
| johnsom | jiteka So in octavia it says "BOOTING"? | 17:44 |
| *** sapd has quit IRC | 17:44 | |
| johnsom | I mean "BOOTING" only waits a bit over a minute before going to ERROR with the default settings. | 17:48 |
| johnsom | Since nova sets it to active as soon as the process starts. | 17:48 |
| *** rraja has joined #openstack-lbaas | 17:49 | |
| *** AlexStaf has joined #openstack-lbaas | 17:51 | |
| *** blake has joined #openstack-lbaas | 17:59 | |
| openstackgerrit | Michael Johnson proposed openstack/neutron-lbaas master: Neutron-LBaaS to Octavia migration tool https://review.openstack.org/578942 | 18:12 |
| *** abaindur has joined #openstack-lbaas | 18:23 | |
| *** abaindur has quit IRC | 18:46 | |
| *** abaindur has joined #openstack-lbaas | 18:46 | |
| *** atoth has quit IRC | 18:53 | |
| *** blake has quit IRC | 19:01 | |
| *** blake has joined #openstack-lbaas | 19:02 | |
| *** blake has quit IRC | 19:06 | |
| openstackgerrit | Michael Johnson proposed openstack/neutron-lbaas master: Neutron-LBaaS to Octavia migration tool https://review.openstack.org/578942 | 19:14 |
| *** blake has joined #openstack-lbaas | 19:16 | |
| *** blake has quit IRC | 19:20 | |
| *** amuller has quit IRC | 19:24 | |
| *** rraja has quit IRC | 19:31 | |
| *** blake has joined #openstack-lbaas | 19:37 | |
| *** abaindur has quit IRC | 19:45 | |
| *** kobis has joined #openstack-lbaas | 19:48 | |
| kobis | general question regarding driver API: so the driver allocates the VIP port (OK I can allocate myself if I choose), but I must cleanup within the LB delete method? | 19:53 |
| kobis | Is there a missing piece in the driver patch? or is that by design? | 19:53 |
| *** AlexStaf has quit IRC | 19:54 | |
| johnsom | Hmm, interesting question. | 19:55 |
| johnsom | Our driver deletes the VIP when it is deleting the LB resources. | 19:55 |
| kobis | in that case, would it make sense to add a delete callback, which in the amphora case will do nothing? | 19:56 |
| kobis | and if that's not implemented, driver_lib will cleanup? | 19:56 |
| johnsom | Or should we just hook the status update callback when you set the LB to DELETED we make sure the VIP resources are cleaned up. | 19:57 |
| kobis | that makes sense even more | 19:57 |
| johnsom | Yeah, that way it's a double check that resource gets cleaned up | 19:58 |
| kobis | cool, so i'll comment on the patch just so we have that on record | 19:58 |
| johnsom | Perfect, I might be able to add that today | 19:58 |
| kobis | great :) | 19:59 |
| *** abaindur has joined #openstack-lbaas | 20:04 | |
| johnsom | kobis I have another issue to run by you. If the user specifies a vip address that is already in use, right now it bubbles up as a 500 driver error. I was considering adding another driver exception for that case. Any concerns? | 20:05 |
| johnsom | That assumes the driver is creating the vip | 20:05 |
| kobis | Since exceptions are the only way that openstack client shows comprehensive error messages, I think that it makes a lot of sense to have higher exception granularity than "driver error" | 20:08 |
| johnsom | Yeah, the client will pass through the user string, but still a 500 is not the right answer for a user mistake | 20:08 |
| johnsom | kobis Also, on the delete thing. | 20:08 |
| johnsom | I am thinking maybe we should just delete the VIP port on a "DELETED" call back if octavia allocated the VIP. That way if the driver created the VIP and wants to reuse it in some way, they would have the option. Thoughts? | 20:09 |
| kobis | I don't see why the driver would want that, but it makes sense that whoever created a resource, is responsible to cleanup | 20:10 |
| johnsom | Yeah, I think that will be my plan. Are you doing port allocation or are you leaving it to Octavia? | 20:11 |
| kobis | I leave it to Octavia. NLBaaS did it, so we try to stick to the old code :) | 20:12 |
| johnsom | Ok, sounds good. Thanks! | 20:12 |
| johnsom | Hopefully you folks are pretty close to done??? | 20:12 |
| kobis | Yeah we have a working yet untested driver :) | 20:12 |
| kobis | which means that our QE will find it broken in various ways | 20:13 |
| johnsom | Nice. We do have a decent tempest plugin if you want.... | 20:13 |
| kobis | And I'm trying to think of a CI which will cover that as well | 20:13 |
| kobis | Yeah once patches will merge in Octavia and on our end, I'll start looking into tempest | 20:14 |
| kobis | any other vendor drivers in progress? | 20:16 |
| johnsom | I don't think so, one had a staffing change so delayed | 20:17 |
| johnsom | kobis Don't forget to update this page when you are ready: https://docs.openstack.org/octavia/latest/admin/providers.html | 20:18 |
| kobis | Sure. But would it make sense to update before patches are merged in Octavia and vmware-nsx? | 20:19 |
| johnsom | No, probably not. It's there when you are ready... | 20:19 |
| kobis | OK | 20:19 |
| xgerman_ | mmh, this privsep is getting sillier and sillier | 20:30 |
| xgerman_ | so we fork the privsep server (A) as root, which then start a reciever thread (B) as a root again (why?), and then we start gunicorn, which forks a worker process, which successfully sends stuff to A, A sends stuff back, but B never recieves is and bloxks | 20:32 |
| xgerman_ | our worker process is shedding privileges | 20:32 |
| johnsom | Yeah, the worker in gunicorn should shed privs, at least that is the goarl | 20:33 |
| xgerman_ | yeah, I have that | 20:33 |
| xgerman_ | as I said it’s getting the results back where the whole thing bombs | 20:34 |
| xgerman_ | I *would* think the reciever process should get started under th worker | 20:35 |
| xgerman_ | or thread | 20:35 |
| xgerman_ | will pester the oslo people | 20:35 |
| johnsom | +1 | 20:36 |
| *** abaindur_ has joined #openstack-lbaas | 20:36 | |
| *** abaindur has quit IRC | 20:36 | |
| *** abaindur has joined #openstack-lbaas | 20:40 | |
| xgerman_ | my money is that the lock doesn’t connect properly aka why would a process running with less priviliges be allowed to read the memory of a more priviliged process… | 20:42 |
| *** abaindur_ has quit IRC | 20:42 | |
| *** abaindur_ has joined #openstack-lbaas | 20:42 | |
| *** kobis has quit IRC | 20:43 | |
| *** abaindur has quit IRC | 20:45 | |
| *** abaindur has joined #openstack-lbaas | 20:46 | |
| *** abaindu__ has joined #openstack-lbaas | 20:47 | |
| *** abaindur_ has quit IRC | 20:48 | |
| *** abaindur has quit IRC | 20:50 | |
| *** abaindu__ has quit IRC | 20:53 | |
| *** abaindur has joined #openstack-lbaas | 21:02 | |
| openstackgerrit | Michael Johnson proposed openstack/octavia master: Fixes unlimited listener connection limit https://review.openstack.org/580724 | 21:27 |
| *** abaindur has quit IRC | 21:32 | |
| *** abaindur has joined #openstack-lbaas | 21:33 | |
| openstackgerrit | Michael Johnson proposed openstack/neutron-lbaas master: Neutron-LBaaS to Octavia migration tool https://review.openstack.org/578942 | 21:37 |
| nmagnezi | johnsom, o/ | 21:44 |
| johnsom | nmagnezi o/ | 21:44 |
| nmagnezi | johnsom, that local.conf for the migration tool devstack | 21:44 |
| nmagnezi | johnsom, can I have it ? :D | 21:45 |
| nmagnezi | I want to prep a node for testing | 21:45 |
| johnsom | nmagnezi This is the one the gate is running: http://logs.openstack.org/42/578942/50/check/neutron-lbaas-to-octavia-migration/330d126/controller/logs/local_conf.txt | 21:46 |
| johnsom | But I can give you a simpler one if you want | 21:46 |
| nmagnezi | Please :) | 21:47 |
| nmagnezi | I can clean it up but.. if you already have it :) | 21:47 |
| johnsom | http://paste.openstack.org/show/725371/ | 21:48 |
| johnsom | That is the one I am using local | 21:48 |
| *** abaindur_ has joined #openstack-lbaas | 21:49 | |
| johnsom | The namespace provider is "haproxy" and the default is octavia | 21:49 |
| *** abaindur has quit IRC | 21:53 | |
| lxkong | johnsom: thanks so much for the patch! | 21:57 |
| johnsom | lxkong Sure, sorry for the trouble! | 21:57 |
| lxkong | johnsom: one more question, does octavia know how many active connections are there in a member? | 21:57 |
| lxkong | we are still thinking about how to provide connection draining feature doc to our customers | 21:58 |
| openstackgerrit | Michael Johnson proposed openstack/octavia master: Fixes unlimited listener connection limit https://review.openstack.org/580724 | 21:58 |
| lxkong | if that's thing like you said the customer should do that outside of octavia(using ansible or something else), they need to know when is the good time to remove the member | 21:58 |
| johnsom | lxkong Yes, I think it does. One minute | 21:59 |
| lxkong | johnsom: sure, thanks for checking | 21:59 |
| *** AlexStaf has joined #openstack-lbaas | 21:59 | |
| johnsom | yes, we get that metric in the stats socket. We don't currently collect that information however. | 22:03 |
| *** abaindur has joined #openstack-lbaas | 22:06 | |
| *** abaindur_ has quit IRC | 22:07 | |
| lxkong | johnsom: you mean here https://github.com/openstack/octavia/blob/8f3eeb5b2e1ddfdca8e22d04dae2d892c9bcef47/octavia/amphorae/backends/utils/haproxy_query.py#L77? This function is collecting everything including the member stats? | 22:10 |
| lxkong | but in here https://github.com/openstack/octavia/blob/8f3eeb5b2e1ddfdca8e22d04dae2d892c9bcef47/octavia/amphorae/backends/health_daemon/health_daemon.py#L133 we only care about `row['svname'] == 'FRONTEND'` | 22:11 |
| johnsom | lxkong yes. | 22:11 |
| johnsom | lxkong This is the output of the socket: | 22:12 |
| johnsom | http://paste.openstack.org/show/725376/ | 22:12 |
| johnsom | Where 7ce0fb2d-fee6-45e5-9e3a-b8fdada3dea6 and a786a26a-019d-4791-a430-b52961db8d58 are members | 22:12 |
| johnsom | The stot column is the connection count per for the member | 22:13 |
| *** numans has quit IRC | 22:14 | |
| johnsom | lxkong The bummer here is we don't really want to collect stats per member and store them in the DB as that could be a lot of overhead | 22:14 |
| lxkong | johnsom: ok, we have the data source, just ignored. But back to the feature, I do think it will be more convenient if octavia could provide a connection draining api on members. How about we evaluate the connection number on the fly after the api is called? | 22:15 |
| johnsom | lxkong I think we would need to make this an on-demand query | 22:15 |
| lxkong | for `on-demand query`, you mean expose an API within the amphora? | 22:16 |
| *** blake has quit IRC | 22:16 | |
| johnsom | lxkong I guess how I would do it is add /v2.0/lbaas/pools/{pool_id}/members/{member-id}/stats to the API | 22:17 |
| lxkong | yeah, that's one option | 22:17 |
| johnsom | lxkong Then when called, make a call out to the amphora to query the current stot. We don't want to do the other stats as failover would impact the accuracy of the other status | 22:17 |
| johnsom | FYI, here are the column descriptions: http://cbonte.github.io/haproxy-dconv/1.6/management.html#9.1 | 22:18 |
| lxkong | thanks for the link | 22:19 |
| *** numans has joined #openstack-lbaas | 22:19 | |
| *** fnaval has quit IRC | 22:19 | |
| johnsom | lxkong opps, it's "scur" that you want | 22:19 |
| *** rcernin has joined #openstack-lbaas | 22:20 | |
| johnsom | lxkong Any of the cumulative columns will be inaccurate as they get reset on any VRRP or amp failover. | 22:20 |
| *** threestrands_ has joined #openstack-lbaas | 22:20 | |
| *** threestrands_ has quit IRC | 22:20 | |
| *** threestrands_ has joined #openstack-lbaas | 22:20 | |
| *** abaindur_ has joined #openstack-lbaas | 22:22 | |
| lxkong | yeah, that would be the restrict. but when lb is in failover, we can not set weight on members, so no need to check the stats actually, that's ok for the connection draining process | 22:23 |
| *** jappleii__ has joined #openstack-lbaas | 22:23 | |
| johnsom | lxkong I mean after the failover completes, the values reset. So as long as you stick to only showing metrics that are "current" and not cumulative you will be fine. | 22:23 |
| johnsom | lxkong so "scur" would be fine. | 22:24 |
| lxkong | ahh, yeah, thanks for the explanation | 22:24 |
| *** abaindur has quit IRC | 22:24 | |
| lxkong | that's what we care about | 22:24 |
| johnsom | +1. | 22:24 |
| *** jappleii__ has quit IRC | 22:24 | |
| *** jappleii__ has joined #openstack-lbaas | 22:25 | |
| lxkong | johnsom: may i know why is it unlikely that we support a single api for that? | 22:25 |
| johnsom | lxkong Not sure I follow, what do you mean by single API? like have it part of member status? | 22:26 |
| *** threestrands_ has quit IRC | 22:26 | |
| lxkong | johnsom: something like `/v2.0/lbaas/pools/{pool_id}/members/{member-id}/drain` | 22:27 |
| johnsom | lxkong Wouldn't that be duplicate? We already manage the weight via the member PUT API | 22:28 |
| johnsom | We might as well have stats in case we later add some of these other "current" stats | 22:28 |
| *** abaindur has joined #openstack-lbaas | 22:28 | |
| johnsom | Plus it is consistent with the listener and load balancer /stats path | 22:28 |
| *** abaindur_ has quit IRC | 22:30 | |
| lxkong | yeah, i understand. I am thinking if we could make the process more easy for the user. For now, the process will be: 1. set weight 0 on the member; 2. watch the member status(once we support it) or wait for a timeout; 3. remove the member form the pool | 22:30 |
| johnsom | Yes, that is the process | 22:30 |
| johnsom | That is the same process you go through with hardware load balancers | 22:31 |
| lxkong | it's acceptable, i just want to know the design here :-) | 22:31 |
| lxkong | `That is the same process you go through with hardware load balancers`, i got the answer now | 22:32 |
| johnsom | I am open to other ideas. I just have the perspective of trying to keep it consistent | 22:32 |
| *** KeithMnemonic has joined #openstack-lbaas | 22:32 | |
| lxkong | johnsom: i will create a story to track this feature request. | 22:32 |
| johnsom | +1 | 22:32 |
| lxkong | johnsom: thanks for your answers. I will test your patch to create a fedora amphora image at the same time. | 22:33 |
| johnsom | cgoncalves http://logs.openstack.org/73/581073/1/check/octavia-v1-dsvm-scenario-kvm-centos.7/523530e/logs/devstacklog.txt.gz#_2018-07-09_16_42_41_847 | 22:35 |
| johnsom | It looks like fedora uses dnf which failed on a remove of those packages, but the centos yum returned 0 | 22:36 |
| lxkong | yeah, centos is ok | 22:37 |
| johnsom | So remove a package with DNF that doesn't exist is an error, but with yum removing a non-existent package is fine. | 22:37 |
| johnsom | That is why the centos gate didn't catch it. | 22:37 |
| cgoncalves | interesting! yum has different behaviors in fedora and centos | 22:39 |
| johnsom | cgoncalves Fedora is using dnf | 22:39 |
| johnsom | I don't know if that is a DIB choice or other.... | 22:39 |
| cgoncalves | http://paste.openstack.org/show/725377/ | 22:40 |
| johnsom | cgoncalves lol, oye. I gave it the benefit of the doubt, but yeah, sad..... | 22:40 |
| *** abaindur_ has joined #openstack-lbaas | 22:45 | |
| xgerman_ | ok, this is the part where privsep is hanging: https://github.com/openstack/oslo.privsep/blob/master/oslo_privsep/comm.py#L134 | 22:46 |
| *** abaindur has quit IRC | 22:47 | |
| cgoncalves | perhaps a 'ignore_nonexistent: true|false' option in packages-install would be worth adding...? | 22:47 |
| openstackgerrit | Michael Johnson proposed openstack/neutron-lbaas master: Neutron-LBaaS to Octavia migration tool https://review.openstack.org/578942 | 22:52 |
| johnsom | Ok, that seems to be passing the migration. I had just disabled the session persistence to figure out all the bugs in the namespace driver. This run should pass, then I will do one more commit changing it to a periodic job. | 22:53 |
| cgoncalves | excellent! really great job! | 22:55 |
| *** abaindur has joined #openstack-lbaas | 22:57 | |
| *** abaindur_ has quit IRC | 23:00 | |
| johnsom | Yeah, a weeks work and 50 revisions to prove what I already tested local... Ugh. Well, done now | 23:10 |
| johnsom | lxkong I just had a few more thoughts about the design to get that "current connections" metric per-member. | 23:11 |
| johnsom | We don't have good way to do that right now with the driver model..... | 23:11 |
| lxkong | why? i'm not familiar with the current driver model | 23:12 |
| johnsom | I mean the driver side isn't bad, it's how it's handled in the driver | 23:12 |
| johnsom | Right now the API processes don't have access to the amphora API. Everything goes over a message queue out to the workers.... | 23:13 |
| johnsom | Ugh, this is really where I wish I could take the time to re-write the status/stats handling code. | 23:14 |
| johnsom | Yeah, so, ok, I guess we have to take the DB hit and pass it with the rest of the stats | 23:14 |
| lxkong | johnsom: i don't understand why we can not send reqeust to o-worker then the amphora to get the value? | 23:17 |
| johnsom | It would be introducing a synchronous call across the queue which we don't do today. | 23:18 |
| lxkong | johnsom: do you mean that all the reqeusts the o-worker is handling are all asynchronous? | 23:18 |
| johnsom | lxkong Correct, today everything the o-worker does is async | 23:19 |
| lxkong | hmm... | 23:19 |
| johnsom | We could add that, but we might be tipping over to just adding it to the DB for now is good enough..... | 23:19 |
| lxkong | johnsom: ok, fair enough. Then we could get the information as easy as getting the lb status | 23:21 |
| johnsom | Yeah, it is just additional info in the message | 23:21 |
| *** ianychoi_ has joined #openstack-lbaas | 23:21 | |
| *** jlaffaye_ has joined #openstack-lbaas | 23:22 | |
| *** jlaffaye_ has quit IRC | 23:22 | |
| *** jlaffaye_ has joined #openstack-lbaas | 23:22 | |
| *** jlaffaye has quit IRC | 23:22 | |
| *** ianychoi has quit IRC | 23:25 | |
| lxkong | johnsom: are you able to help do that? | 23:26 |
| xgerman_ | ok, solved the privsep myster —I need to run the privsep piece as part of the worker | 23:26 |
| johnsom | lxkong No, it would need to be a Stein activity if I do it. We have too much going on for Rocky already IMO. | 23:27 |
| johnsom | Stein we could probably get it in pretty quick | 23:27 |
| *** abaindur has quit IRC | 23:30 | |
| lxkong | johnsom: ok, good to know. thanks! | 23:35 |
| *** abaindur has joined #openstack-lbaas | 23:37 | |
| johnsom | We are just three weeks from feature freeze on Rocky... | 23:37 |
| *** abaindur has quit IRC | 23:50 | |
| *** abaindur has joined #openstack-lbaas | 23:53 | |
Generated by irclog2html.py 2.15.3 by Marius Gedminas - find it at mg.pov.lt!