| rm_work | johnsom: i'm just not seeing how this patch helps unless "fixed_ips" contains more than one address, and I'm not sure how it would? | 00:12 |
|---|---|---|
| rm_work | wait, *is this* inside the amp? or not | 00:14 |
| openstackgerrit | Michael Johnson proposed openstack/octavia master: Adds flavor support to the amphora driver https://review.openstack.org/621323 | 00:14 |
| rm_work | ah yeah ok it is | 00:15 |
| johnsom | rm_work It's inside | 00:15 |
| rm_work | naming collision with something we use outside confused me | 00:15 |
| rm_work | so we pass the port_info into the request to the amp | 00:16 |
| rm_work | which includes the fixed_ips list | 00:16 |
| rm_work | but I'm not sure what case we'd generate that with multiple IPs | 00:16 |
| johnsom | Nova tells us that list when we get the port back | 00:16 |
| rm_work | right ok | 00:17 |
| rm_work | but this goes back to "how does nova know what subnet we want" | 00:17 |
| rm_work | and why would nova ever give us a port with multiple | 00:17 |
| johnsom | I don't think it does, I think it sends them all in the way we are asking for stuff. | 00:17 |
| rm_work | so this takes care of the case of them not coming up right internally, assuming it's set up properly externally | 00:17 |
| rm_work | but we would never add the right subnets to the ports | 00:18 |
| johnsom | right, basically we were getting a list len > 1 and we were only setting up one address | 00:18 |
| rm_work | because i think it's pure dumb luck this works at all | 00:18 |
| johnsom | That patch makes it bring them all up | 00:18 |
| rm_work | because per earlier, we appear to just ask nova for a port on a network | 00:18 |
| rm_work | and it *guesses* what subnet | 00:18 |
| rm_work | because we never specify that | 00:18 |
| johnsom | No, neutron will give ALL subnets on the network back in the port details | 00:18 |
| rm_work | errr, that's not true | 00:19 |
| rm_work | i'm looking at one right now | 00:19 |
| johnsom | If we specify it would only be one | 00:19 |
| rm_work | if i create a port on a network in neutron | 00:19 |
| rm_work | hmmmm ok | 00:20 |
| rm_work | so it does | 00:20 |
| rm_work | if i just "create a port" | 00:20 |
| rm_work | but i'm looking at ports on LBs | 00:20 |
| rm_work | oh, maybe it also does? | 00:20 |
| rm_work | one sec | 00:20 |
| rm_work | hmm no | 00:21 |
| rm_work | err wtf | 00:21 |
| rm_work | ok give me a minute | 00:21 |
| rm_work | the ports aren't actually bound to the server? | 00:22 |
| johnsom | https://www.irccloud.com/pastebin/q7PpjFFK/ | 00:23 |
| rm_work | ok that's not what I get | 00:24 |
| rm_work | http://paste.openstack.org/show/UcqKSu5lHm6SEDZFSd72/ | 00:25 |
| rm_work | that's one of the two subnets on the network | 00:25 |
| rm_work | it didn't bind the ipv4 subnet on that port | 00:26 |
| rm_work | do i need the patch for that to happen correctly? | 00:26 |
| johnsom | No, that is master + my flavors stuff, but not the IPv6 fixes | 00:27 |
| rm_work | hmmm | 00:27 |
| johnsom | It has worked that way for me for years frankly | 00:27 |
| rm_work | so on devstack, the private network has two subnets, right? | 00:27 |
| johnsom | Mine does as I'm IPv6 enabled on the host | 00:27 |
| rm_work | private-subnet and ipv6-private-subnet | 00:27 |
| rm_work | yes | 00:27 |
| rm_work | so | 00:28 |
| johnsom | yeah | 00:28 |
| rm_work | i'm creating a new LB | 00:28 |
| johnsom | I do: openstack loadbalancer create --name lb1 --vip-subnet-id private-subnet | 00:28 |
| johnsom | An that is what created the port I pasted you | 00:28 |
| rm_work | right ok | 00:28 |
| rm_work | so i just did that | 00:28 |
| rm_work | waiting for it to come up | 00:29 |
| *** yamamoto has quit IRC | 00:49 | |
| rm_work | hmmmmmmm | 00:51 |
| rm_work | so when i make an LB with ipv4 VIP, it does like yours | 00:52 |
| *** PagliaccisCloud has joined #openstack-lbaas | 00:52 | |
| rm_work | when I make it with an ipv6 VIP... | 00:52 |
| rm_work | the ports only get ipv6, no ipv4 | 00:52 |
| rm_work | SO | 00:52 |
| rm_work | when I create a member with an ipv6 subnet specified, it creates the port with only ipv6 | 00:53 |
| rm_work | and then when i add a member with ipv4, it doesn't do anything with the port, and there's no ipv4 routes | 00:53 |
| rm_work | but i bet if i do it the other way around, it'll be ok? | 00:53 |
| openstackgerrit | Michael Johnson proposed openstack/octavia master: Add provider driver capabilities API https://review.openstack.org/624190 | 00:54 |
| openstackgerrit | Michael Johnson proposed openstack/octavia master: Add provider driver capabilities API https://review.openstack.org/624190 | 00:54 |
| rm_work | ahhhh yeah wow even with the VIP, when I tell it to make one on private-subnet, i get an ipv4 and ipv6 VIP O_o | 00:59 |
| rm_work | but if i tell it to make it on the ipv6-private-subnet it does not get both | 01:01 |
| rm_work | must be a neutron peculiarity | 01:01 |
| rm_work | they assume "you always get ipv6 no matter what" | 01:02 |
| johnsom | I know we have some funky code for "picking" subnets/addresses for the VIP | 01:05 |
| rm_work | hmmm ahhh maybe i'm having issues because of that | 01:06 |
| rm_work | yeah hah i bet | 01:06 |
| rm_work | one sec | 01:06 |
| rm_work | ok yeah | 01:12 |
| rm_work | so your patch fixes it | 01:13 |
| rm_work | the issue *I* was having was specifically because I was doing member on vip network | 01:13 |
| rm_work | and i was using an ipv6 VIP | 01:13 |
| rm_work | which then didn't allow an ipv4 member | 01:13 |
| rm_work | >_< | 01:13 |
| rm_work | but if i use your patch and do an ipv6 vip on a DIFFERENT subnet from my members, it works fine | 01:14 |
| rm_work | johnsom: were you maintaining that patch or no | 01:14 |
| rm_work | johnsom: i'd like to merge it, but it needs a test or something... | 01:14 |
| johnsom | I moved on and will eventually circle back, but feel free to take it and run | 01:14 |
| rm_work | k | 01:15 |
| rm_work | i'l see if i can figure out where to add an appropriate test | 01:15 |
| rm_work | ugh i want to submit a fix for the test i was debugging earlier even if i won't be updating the code lol | 01:15 |
| openstackgerrit | Adam Harwell proposed openstack/octavia master: demystify the calculate_network_delta unit test https://review.openstack.org/624259 | 01:18 |
| openstackgerrit | Michael Johnson proposed openstack/octavia master: Add provider driver capabilities API https://review.openstack.org/624190 | 01:18 |
| *** ianychoi has quit IRC | 01:20 | |
| openstackgerrit | Michael Johnson proposed openstack/octavia master: Adds flavor support to the amphora driver https://review.openstack.org/621323 | 01:24 |
| openstackgerrit | Michael Johnson proposed openstack/octavia master: Add provider driver capabilities API https://review.openstack.org/624190 | 01:24 |
| rm_work | ok, tomorrow | 01:28 |
| rm_work | i'll figure that out | 01:28 |
| openstackgerrit | ZhaoBo proposed openstack/octavia master: Add boolean TLS option into Pool for backend re-encryption https://review.openstack.org/624264 | 03:08 |
| openstackgerrit | ZhaoBo proposed openstack/python-octaviaclient master: Add TLS option into Pool CLI https://review.openstack.org/624265 | 03:09 |
| *** hongbin has joined #openstack-lbaas | 03:15 | |
| openstackgerrit | ZhaoBo proposed openstack/octavia master: WIP: Add boolean TLS option into Pool for backend re-encryption https://review.openstack.org/624264 | 03:26 |
| openstackgerrit | ZhaoBo proposed openstack/octavia master: Amp driver support sni option to send the hostname to backend https://review.openstack.org/624267 | 03:26 |
| *** bzhao__ has joined #openstack-lbaas | 03:28 | |
| bzhao__ | LOL. ;-) | 03:51 |
| openstackgerrit | ZhaoBo proposed openstack/octavia master: Add boolean TLS option into Pool for backend re-encryption https://review.openstack.org/624264 | 03:54 |
| *** ramishra has quit IRC | 03:59 | |
| *** yamamoto has joined #openstack-lbaas | 04:46 | |
| *** jarodwl has joined #openstack-lbaas | 05:02 | |
| *** hongbin has quit IRC | 05:07 | |
| *** ramishra has joined #openstack-lbaas | 05:26 | |
| *** yboaron_ has joined #openstack-lbaas | 06:00 | |
| *** jmccrory has quit IRC | 06:34 | |
| *** jmccrory has joined #openstack-lbaas | 06:40 | |
| *** rcernin has quit IRC | 06:43 | |
| openstackgerrit | Yang JianFeng proposed openstack/python-octaviaclient master: Add provider command support https://review.openstack.org/624294 | 07:00 |
| openstackgerrit | xulei proposed openstack/octavia-dashboard master: Modify http to https. https://review.openstack.org/624296 | 07:08 |
| openstackgerrit | xulei proposed openstack/octavia-dashboard master: Modify http to https. https://review.openstack.org/624296 | 07:10 |
| *** ramishra has quit IRC | 07:15 | |
| *** ccamposr has joined #openstack-lbaas | 07:34 | |
| *** yboaron_ has quit IRC | 07:35 | |
| openstackgerrit | Yang JianFeng proposed openstack/python-octaviaclient master: Add provider command support https://review.openstack.org/624294 | 07:40 |
| *** oanson has quit IRC | 07:54 | |
| *** ramishra has joined #openstack-lbaas | 08:01 | |
| *** rpittau has joined #openstack-lbaas | 08:05 | |
| *** gcheresh_ has joined #openstack-lbaas | 08:08 | |
| *** rpittau has quit IRC | 08:09 | |
| openstackgerrit | Yang JianFeng proposed openstack/python-octaviaclient master: Add provider command support https://review.openstack.org/624294 | 08:11 |
| *** velizarx has joined #openstack-lbaas | 08:12 | |
| openstackgerrit | Reedip proposed openstack/octavia-tempest-plugin master: Modify Member tests for Provider Drivers https://review.openstack.org/598476 | 08:19 |
| *** rpittau has joined #openstack-lbaas | 08:26 | |
| *** ramishra has quit IRC | 08:44 | |
| *** ramishra has joined #openstack-lbaas | 08:51 | |
| *** yamamoto has quit IRC | 09:01 | |
| *** yamamoto has joined #openstack-lbaas | 09:18 | |
| *** rpittau has quit IRC | 09:24 | |
| *** yamamoto has quit IRC | 09:27 | |
| *** pbourke_ has quit IRC | 09:54 | |
| *** yamamoto has joined #openstack-lbaas | 10:06 | |
| *** salmankhan has joined #openstack-lbaas | 10:08 | |
| *** yamamoto has quit IRC | 10:10 | |
| *** salmankhan has quit IRC | 10:21 | |
| openstackgerrit | Carlos Goncalves proposed openstack/python-octaviaclient master: Add failover an amphora client support https://review.openstack.org/532424 | 10:21 |
| *** salmankhan has joined #openstack-lbaas | 10:21 | |
| *** rpittau has joined #openstack-lbaas | 10:26 | |
| *** yamamoto has joined #openstack-lbaas | 10:45 | |
| *** yamamoto has quit IRC | 10:47 | |
| *** tobias-urdin is now known as tobias-urdin|lun | 11:00 | |
| *** tobias-urdin|lun is now known as tobias-urdin_afk | 11:01 | |
| *** yamamoto has joined #openstack-lbaas | 11:11 | |
| *** yamamoto has quit IRC | 11:16 | |
| *** yamamoto has joined #openstack-lbaas | 11:16 | |
| *** tobias-urdin_afk is now known as tobias-urdin | 11:27 | |
| *** salmankhan has quit IRC | 11:46 | |
| *** salmankhan has joined #openstack-lbaas | 11:50 | |
| *** yamamoto has quit IRC | 12:23 | |
| *** rpittau is now known as rpittau|lunch | 12:25 | |
| *** yamamoto has joined #openstack-lbaas | 12:28 | |
| *** yamamoto has quit IRC | 12:32 | |
| openstackgerrit | Merged openstack/octavia master: Correct the base class for v2 l7rules type tests https://review.openstack.org/619397 | 12:35 |
| *** velizarx has quit IRC | 12:49 | |
| *** velizarx has joined #openstack-lbaas | 13:03 | |
| *** yamamoto has joined #openstack-lbaas | 13:05 | |
| *** rpittau|lunch is now known as rpittau | 13:31 | |
| *** aojea_ has joined #openstack-lbaas | 13:45 | |
| *** reedipb has joined #openstack-lbaas | 13:59 | |
| *** salmankhan has quit IRC | 14:01 | |
| *** salmankhan has joined #openstack-lbaas | 14:01 | |
| reedipb | @johnsom, @xgerman , @cgoncalves : hi | 14:02 |
| reedipb | @johnsom, @xgerman , @cgoncalves : I wanted to know a bit about the flavors concept in Octavia | 14:02 |
| reedipb | @johnsom, @xgerman , @cgoncalves : precisely my question is , if the flavors functionality for Octavia Provider drivers is enabled, and we have both Amphorae ( which supports L7) and OVN Provider Driver( which supports L4) , and then if the user sends a request to create an HTTP Listener, how would the operation occur? | 14:07 |
| openstackgerrit | Arnaud Morin proposed openstack/octavia master: Update requirements for ubuntu https://review.openstack.org/624405 | 14:08 |
| *** aojea_ has quit IRC | 14:14 | |
| openstackgerrit | Arnaud Morin proposed openstack/octavia master: Update requirements for ubuntu https://review.openstack.org/624405 | 14:15 |
| *** salmankhan has quit IRC | 14:46 | |
| cgoncalves | reedipb, if the LB is provided by the amphora driver, L7 listener creation request will succeed. if by OVN provider driver, it will fail | 14:51 |
| *** velizarx has quit IRC | 14:53 | |
| reedipb | cgoncalves: thanks for the info. I am not sure about the default behavior ( that would be based on implementation of the flavors functionality, I guess). Can you let me know if the flavors functionality is targetted in T release ? or later? | 14:56 |
| *** salmankhan has joined #openstack-lbaas | 14:59 | |
| cgoncalves | reedipb, Michael has started flavor implementation last week or so. the expected plan is to have basic flavor support in Stein | 15:05 |
| *** gcheresh_ has quit IRC | 15:17 | |
| *** oanson has joined #openstack-lbaas | 15:17 | |
| *** salmankhan1 has joined #openstack-lbaas | 15:20 | |
| *** salmankhan has quit IRC | 15:20 | |
| *** salmankhan1 is now known as salmankhan | 15:20 | |
| johnsom | reedipb: flavors are defined per provider, so the driver has to support them or the settings will be the defaults defined in the octavia.conf. Flavors are a Stein feature and most of the base work is done. I just need to add more settings, docs, client support, and tempest. | 15:26 |
| *** aojea_ has joined #openstack-lbaas | 15:36 | |
| *** ivve has joined #openstack-lbaas | 15:37 | |
| *** gcheresh_ has joined #openstack-lbaas | 15:40 | |
| xgerman | reedipb: I would assume the system pick the provider based on the flavor | 15:50 |
| openstackgerrit | Arnaud Morin proposed openstack/octavia master: Update requirements for ubuntu https://review.openstack.org/624405 | 15:51 |
| *** gcheresh_ has quit IRC | 15:53 | |
| openstackgerrit | Michael Johnson proposed openstack/octavia master: Make the CentOS 7 scenario gate non-voting https://review.openstack.org/623070 | 16:12 |
| cgoncalves | ^ understandably :( | 16:15 |
| johnsom | Yeah, we are still seeing failures. Timeout failures to be exact. It looks like the amps aren't booting up in a timely fashion in some runs. | 16:16 |
| openstackgerrit | Michael Johnson proposed openstack/octavia master: Add provider driver capabilities API https://review.openstack.org/624190 | 16:16 |
| brtknr | hmm why does my cloud-controller-manager keep creating multiple loadbalancers for the same service? | 16:17 |
| johnsom | Do you have Octavia configured for Active/Standby HA mode? That would cause two instances to be booted for each load balancer | 16:18 |
| reedipb | thanks johnsom, xgerman, cgoncalves :) | 16:19 |
| openstackgerrit | Michael Johnson proposed openstack/octavia-tempest-plugin master: Adds get_amphora_stats to the service client https://review.openstack.org/585060 | 16:26 |
| *** rpittau has quit IRC | 16:30 | |
| *** aojea_ has quit IRC | 17:02 | |
| *** aojea_ has joined #openstack-lbaas | 17:03 | |
| *** aojea_ has quit IRC | 17:07 | |
| *** yamamoto has quit IRC | 17:19 | |
| *** amuller has joined #openstack-lbaas | 17:48 | |
| johnsom | rm_work Barbican/Octavia question. In queens, when octavia reaches out to barbican to pull a secret, it is using the "octavia" service account right? Not impersonating the user? I am seeing strange barbican policy failure ERRORs even though I have added the "octavia" user to every RBAC role I can think of.... | 17:49 |
| *** salmankhan has quit IRC | 17:51 | |
| *** Swami has joined #openstack-lbaas | 18:09 | |
| openstackgerrit | Luis Tomas Bolivar proposed openstack/octavia master: Enabling SG customization on loadbalancer listerners https://review.openstack.org/602564 | 18:15 |
| openstackgerrit | Luis Tomas Bolivar proposed openstack/octavia master: Enabling SG customization on loadbalancer listerners https://review.openstack.org/602564 | 18:18 |
| jiteka | I noticed that when creating a LB, it comes with a dedicated secgroup named lb-<lb uuid> with 2 rules associated to it | 18:21 |
| jiteka | Is it a default logic or due to configuration/misconfiguration on my side | 18:21 |
| jiteka | Right now I'm facing this error : ERROR oslo_messaging.rpc.server [req-4964905f-3ac1-4e38-9060-e2ee561cea8d - 57bae527dd574db09f46c04286fb0e39 - - -] Exception during message handling: OverQuotaClient: Quota exceeded for resources: ['security_group_rule']. | 18:21 |
| jiteka | At first I was thinking that impacted secgroup quota relate to the project which subnet-id used to create new LB belong to | 18:23 |
| openstackgerrit | Swaminathan Vasudevan proposed openstack/neutron-lbaas stable/ocata: Improve speed of listing from DB https://review.openstack.org/624160 | 18:31 |
| openstackgerrit | Swaminathan Vasudevan proposed openstack/neutron-lbaas stable/pike: Improve speed of listing from DB https://review.openstack.org/624152 | 18:38 |
| johnsom | jiteka Hi, yes, each LB gets a security group that opens the ports the user specifies for the Listeners. | 18:40 |
| johnsom | The account you are running octavia under should have non-default quotas.... | 18:41 |
| johnsom | Here is how German set it up for OSA: https://github.com/openstack/openstack-ansible-os_octavia/blob/master/defaults/main.yml#L473 | 18:42 |
| Swami | johnsom: Should we cherry-pick these two patches together ( Based on the commit message in the later it seems there is some dependency in there), can you confirm. https://review.openstack.org/#/c/624152/ and https://review.openstack.org/#/c/624225/. | 18:43 |
| *** ccamposr has quit IRC | 18:52 | |
| johnsom | Swami Yes, I think you need to. Though there is some controversy over the https://review.openstack.org/#/c/624225 patch | 18:53 |
| johnsom | https://review.openstack.org/#/c/616763/ | 18:53 |
| johnsom | Swami Ocata? Really? | 18:54 |
| Swami | johnsom: Yes I read the controversy issue in that patch, where someone was trying to revert it. | 18:55 |
| Swami | johnsom: Our target is still old, newton. | 18:55 |
| *** aojea has joined #openstack-lbaas | 18:56 | |
| Swami | johnsom: Ok I will cherry-pick both the patches then. | 18:56 |
| *** yamamoto has joined #openstack-lbaas | 19:17 | |
| openstackgerrit | Swaminathan Vasudevan proposed openstack/neutron-lbaas stable/ocata: Updated "create_pool" method in plugin https://review.openstack.org/624468 | 19:25 |
| openstackgerrit | Swaminathan Vasudevan proposed openstack/neutron-lbaas stable/ocata: Improve speed of listing from DB https://review.openstack.org/624160 | 19:25 |
| *** fnaval has joined #openstack-lbaas | 19:49 | |
| jiteka | thanks johnsom | 19:59 |
| jiteka | johnsom: do you know if there is a way having octavia not making a security group for every new LB and falling back on a default one when listener ports are the same ? It would avoid to get a massive increase of secgroup-rule based on number of active LBs | 20:01 |
| johnsom | Currently we do not have that capability. | 20:02 |
| jiteka | johnsom: ok thanks, I will adapt octavia tenant quota then to fix my problem | 20:06 |
| *** yamamoto has quit IRC | 20:19 | |
| *** aojea has quit IRC | 21:19 | |
| *** amuller has quit IRC | 21:19 | |
| *** rcernin has joined #openstack-lbaas | 21:59 | |
| rm_work | johnsom: should be, yes | 22:07 |
| openstackgerrit | Hongbin Lu proposed openstack/neutron-lbaas master: [DNM] Test neutron policy-in-code https://review.openstack.org/624510 | 22:09 |
| rm_work | jiteka: the secgroup is kinda necessary, there's no way a default one would work for all LBs | 22:11 |
| rm_work | we could try to flyweight it I GUESS? for like... the most common ones | 22:12 |
| rm_work | but that would add a whole lot of complexity :/ | 22:12 |
| *** yamamoto has joined #openstack-lbaas | 22:55 | |
| openstackgerrit | Merged openstack/octavia master: Make the CentOS 7 scenario gate non-voting https://review.openstack.org/623070 | 22:59 |
| openstackgerrit | Adam Harwell proposed openstack/octavia master: Bring up secondary IPs on member networks https://review.openstack.org/611460 | 23:00 |
| rm_work | johnsom: ^^ testing added | 23:00 |
| johnsom | Super, thank you! | 23:00 |
| rm_work | it's in a weird place IMO but that's where it was before and I'm not sure I can think of anywhere better <_< | 23:00 |
| rm_work | in testing my devstack, this worked fine | 23:00 |
| rm_work | so I'm +2 once it passes | 23:01 |
| johnsom | Cool. I am having a bad barbican day... Two folks stuck with issues that I don't have answers for today. | 23:13 |
| johnsom | The morning was Octavia accessing barbican, we got 403 policy errors no matter what I tried for roles on the account. | 23:14 |
| xgerman | yeah, BBQ hates us | 23:15 |
| openstackgerrit | German Eichberger proposed openstack/octavia master: Add auditing support to the Octavia API https://review.openstack.org/609854 | 23:32 |
| xgerman | rebase... | 23:33 |
| openstackgerrit | Merged openstack/octavia-lib master: Initial provider driver library checkin https://review.openstack.org/612815 | 23:41 |
| johnsom | Awesome! We can do a first release of octavia-lib and get it in global-requirements. Super pumped to get that going. | 23:42 |
Generated by irclog2html.py 2.15.3 by Marius Gedminas - find it at mg.pov.lt!