| *** sshank has quit IRC | 00:01 | |
| *** yamamoto_ has joined #openstack-lbaas | 00:18 | |
| *** yamamoto_ has quit IRC | 00:24 | |
| *** catintheroof has quit IRC | 00:25 | |
| *** slaweq has joined #openstack-lbaas | 00:28 | |
| *** slaweq has quit IRC | 00:33 | |
| xgerman_ | I like socket but yeah… | 00:38 |
|---|---|---|
| xgerman_ | also like some reviews on my stuff: proxy, active-active, so I can continuew | 00:39 |
| *** JudeC has quit IRC | 00:50 | |
| rm_work | xgerman_: well I might have to ... looks like gunicorn won't forward signals | 00:55 |
| *** openstackgerrit has joined #openstack-lbaas | 01:12 | |
| openstackgerrit | OpenStack Proposal Bot proposed openstack/octavia master: Updated from global requirements https://review.openstack.org/496052 | 01:12 |
| *** yamamoto_ has joined #openstack-lbaas | 01:21 | |
| *** yamamoto_ has quit IRC | 01:26 | |
| openstackgerrit | Adam Harwell proposed openstack/octavia master: WIP: Floating IP Network Driver (spans L3s) https://review.openstack.org/435612 | 01:34 |
| johnsom | https://usercontent.irccloud-cdn.com/file/ZWbufmf9/1503365957.JPG | 01:39 |
| johnsom | rm_work | 01:39 |
| rm_work | johnsom | 01:39 |
| rm_work | i figured out the signals issue BTW | 01:39 |
| rm_work | and that looks ridiculous | 01:39 |
| rm_work | and now I'm wishing I'd taken you up on your guest room for the eclipse just so I could also be around when you eat that | 01:39 |
| johnsom | Tried my hand at hickory smoking again | 01:40 |
| rm_work | you using a barrel or you have like ... a real smoker | 01:40 |
| rm_work | the barrel method always struck me as more "legit" :P at least if you want oldschool southern bbq | 01:41 |
| rm_work | and has the advantage of being cheap | 01:41 |
| rm_work | but not as easy T_T | 01:41 |
| johnsom | "Real" electric smoker. Parting gift from HP | 01:42 |
| rm_work | nice | 01:44 |
| rm_work | yeah those make it pretty straightforward | 01:44 |
| rm_work | (insomuch as any cooking is straightforward) | 01:44 |
| rm_work | Did it turn out? | 01:45 |
| rm_work | i normally don't eat for a bit but that's making me really hungry T_T | 01:45 |
| rm_work | my favorite now is doing ribs with an Instant-Pot ... which I feel is basically like cheating but are amazing | 01:47 |
| johnsom | It did. My first attempt last year, not so much, but this is pretty good. | 01:48 |
| *** yamamoto_ has joined #openstack-lbaas | 02:22 | |
| *** yamamoto_ has quit IRC | 02:28 | |
| *** catintheroof has joined #openstack-lbaas | 02:30 | |
| *** yamamoto has joined #openstack-lbaas | 02:56 | |
| *** catintheroof has quit IRC | 02:57 | |
| *** yamamoto has quit IRC | 03:02 | |
| *** links has joined #openstack-lbaas | 03:51 | |
| *** yamamoto has joined #openstack-lbaas | 04:03 | |
| *** yamamoto has quit IRC | 04:09 | |
| *** gcheresh has joined #openstack-lbaas | 04:33 | |
| *** gcheresh has quit IRC | 04:36 | |
| *** yamamoto has joined #openstack-lbaas | 05:05 | |
| *** yamamoto has quit IRC | 05:10 | |
| *** gcheresh has joined #openstack-lbaas | 05:21 | |
| *** links has quit IRC | 05:30 | |
| *** links has joined #openstack-lbaas | 05:33 | |
| *** JudeC has joined #openstack-lbaas | 05:46 | |
| rm_work | johnsom: -C for keepalived seems to ... completely make it stop doing anything, including communicating with peers and doing notify scripts :( | 06:00 |
| openstackgerrit | Adam Harwell proposed openstack/octavia master: WIP: Floating IP Network Driver (spans L3s) https://review.openstack.org/435612 | 06:03 |
| *** slaweq has joined #openstack-lbaas | 06:04 | |
| *** yamamoto has joined #openstack-lbaas | 06:06 | |
| openstackgerrit | Adam Harwell proposed openstack/octavia master: WIP: Floating IP Network Driver (spans L3s) https://review.openstack.org/435612 | 06:08 |
| *** yamamoto has quit IRC | 06:12 | |
| *** eezhova has joined #openstack-lbaas | 06:26 | |
| *** armax has quit IRC | 06:39 | |
| *** rcernin has joined #openstack-lbaas | 06:45 | |
| *** pcaruana has joined #openstack-lbaas | 06:45 | |
| *** csomerville has joined #openstack-lbaas | 07:05 | |
| *** cody-somerville has quit IRC | 07:05 | |
| *** eezhova has quit IRC | 07:16 | |
| *** dmellado has joined #openstack-lbaas | 07:18 | |
| *** tesseract has joined #openstack-lbaas | 07:20 | |
| -openstackstatus- NOTICE: Gerrit is going to be restarted due to slow performance | 07:36 | |
| -openstackstatus- NOTICE: Gerrit has been restarted successfully | 07:40 | |
| *** eezhova has joined #openstack-lbaas | 07:45 | |
| *** aojea has joined #openstack-lbaas | 08:32 | |
| *** yamamoto has joined #openstack-lbaas | 08:35 | |
| *** aojea has quit IRC | 08:44 | |
| *** yamamoto has quit IRC | 08:59 | |
| *** yamamoto has joined #openstack-lbaas | 08:59 | |
| *** yamamoto has quit IRC | 08:59 | |
| *** JudeC has quit IRC | 09:00 | |
| *** aojea has joined #openstack-lbaas | 09:12 | |
| isantosp_ | hi, how does loadbalancer quota works? is setted it on a neutron side? | 09:17 |
| *** csomerville has quit IRC | 09:50 | |
| *** yamamoto has joined #openstack-lbaas | 10:00 | |
| *** yamamoto has quit IRC | 10:05 | |
| *** yamamoto has joined #openstack-lbaas | 10:30 | |
| *** atoth has joined #openstack-lbaas | 10:31 | |
| *** slaweq_ has joined #openstack-lbaas | 10:59 | |
| *** junbo has quit IRC | 11:01 | |
| *** slaweq has quit IRC | 11:01 | |
| *** junbo has joined #openstack-lbaas | 11:05 | |
| openstackgerrit | Omer Anson proposed openstack/octavia master: devstack: Allow 3rd party neutron backends deployment https://review.openstack.org/496205 | 11:10 |
| *** yamamoto has quit IRC | 11:18 | |
| *** yamamoto has joined #openstack-lbaas | 11:30 | |
| *** belharar has joined #openstack-lbaas | 11:33 | |
| *** fnaval has quit IRC | 11:41 | |
| *** yamamoto has quit IRC | 11:46 | |
| *** yamamoto has joined #openstack-lbaas | 11:48 | |
| *** yamamoto has quit IRC | 12:03 | |
| *** yamamoto has joined #openstack-lbaas | 12:04 | |
| *** aojea has quit IRC | 12:07 | |
| *** yamamoto has quit IRC | 12:10 | |
| *** aojea has joined #openstack-lbaas | 12:10 | |
| *** yamamoto has joined #openstack-lbaas | 12:12 | |
| *** aojea has quit IRC | 12:21 | |
| *** slaweq_ has quit IRC | 12:25 | |
| *** slaweq has joined #openstack-lbaas | 12:26 | |
| *** slaweq has quit IRC | 12:30 | |
| *** slaweq has joined #openstack-lbaas | 12:34 | |
| *** aojea has joined #openstack-lbaas | 12:36 | |
| *** aojea has quit IRC | 13:05 | |
| *** aojea has joined #openstack-lbaas | 13:10 | |
| *** yamamoto has quit IRC | 13:12 | |
| *** yamamoto has joined #openstack-lbaas | 13:14 | |
| *** yamamoto has quit IRC | 13:22 | |
| *** voelzmo has joined #openstack-lbaas | 13:27 | |
| voelzmo | Hello friends of the loadbalancers! A quick question about the behavior of octavia when adding members: does it add new ports in the subnet where I'm adding members from? | 13:29 |
| *** fnaval has joined #openstack-lbaas | 13:30 | |
| voelzmo | And how does octavia do scaling of the LBaaS? Let's say I'm adding a few thousand members, how does it keep up with the traffic to those instances? | 13:31 |
| *** leitan has joined #openstack-lbaas | 13:34 | |
| johnsom | isantosp_ If you are using neutron-lbaas, yes, it is through neutron. If you are just using the Octavia endpoint quota is part of the API | 13:38 |
| johnsom | voelzmo Yes, we hot plug the networks into the load balancer it they are not already attached. | 13:39 |
| johnsom | At the moment we have not yet finished the act/act work that would help with scaling for the octavia driver. You would need to use a vendor driver for that use case today. | 13:41 |
| voelzmo | @johnsom Thanks! Are there idea on how to implement the scaling part? I'm wondering how the loadbalanced members would identify that the traffic they receive is really coming from the LB and not from somewhere else | 13:42 |
| voelzmo | I'd like to keep security groups as tight as possible | 13:42 |
| voelzmo | @johnsom And a question concerning what you call "hot plugging the network": does that mean octavia creates a new port in that network and all traffic to loadbalanced members will come from that port? | 13:43 |
| johnsom | I see. We have ideas and discussed options, but initially it will use DHCP for the ports | 13:45 |
| johnsom | Yes, that is how it works | 13:45 |
| voelzmo | @johnsom using DHCP would be okay, I'm trying not to care about IP addresses. Currently, we use marker-security-groups which are referenced by the security groups applied to the pool members | 13:50 |
| voelzmo | so they allow e.g. traffic from security group 'lbaas' on port 443 and 22 | 13:51 |
| voelzmo | however, octavia would need to apply the security group to all ports it creates during runtime then | 13:51 |
| *** slaweq has quit IRC | 13:51 | |
| voelzmo | e.g. a user would specify a "loadbalancer security group" when creating an LBaaS and octavia takes care of the rest | 13:51 |
| voelzmo | does that sound reasonable? | 13:51 |
| johnsom | Well, we don't want to blindly apply a user defined security group to the ports. | 13:53 |
| johnsom | Currently you can restrict access to the local subnet. If you are using TLS we plan to add backend encryption with client cert support. | 13:55 |
| *** aojea has quit IRC | 13:55 | |
| *** aojea has joined #openstack-lbaas | 13:56 | |
| voelzmo | @johnsom currently, it would mean that the subnet which contains my members needs to allow for incoming traffic from all IPs in that subnet, correct? | 13:57 |
| voelzmo | There is no way to restrict incoming traffic to only the loadbalancer instances, right? | 13:57 |
| johnsom | I think we need to look at how to layer them in a safe way | 13:57 |
| johnsom | Right, there are some tricks with allocation pools and timing, but not the best solution | 13:58 |
| *** voelzmo has quit IRC | 14:01 | |
| *** bcafarel has quit IRC | 14:05 | |
| *** voelzmo has joined #openstack-lbaas | 14:11 | |
| *** armax has joined #openstack-lbaas | 14:14 | |
| xgerman_ | johnsom we have the disk image build script failing on the ansible side | 14:14 |
| johnsom | Hmm, joy. I will take a look when I get into the office. | 14:15 |
| johnsom | Gates or OSA? | 14:16 |
| voelzmo | @johnsom so you say the problem is known but you wouldn't try to solve this with a user-provided security group which is automatically applied to lbaas-created ports? | 14:17 |
| johnsom | voelzmo We have to do it in a way that doesn't allow the load balancer to be exposed. | 14:18 |
| voelzmo | @johnsom right now I can add manually security groups which expose it, right? | 14:19 |
| johnsom | Right now Octavia managed the port security groups and only opens required ports, but if we take user groups someone could open a hole into the lb | 14:19 |
| johnsom | Only as an Admin, yes | 14:20 |
| *** links has quit IRC | 14:20 | |
| voelzmo | @johnsom really? I've been applying security groups with neutron update-port all day today? | 14:20 |
| *** bcafarel has joined #openstack-lbaas | 14:20 | |
| voelzmo | Or is this just because my vendor-specific backend allows this by placing the ports visibly in my subnet? | 14:21 |
| johnsom | Could be. The octavia ports are owned by Octavia and only Admins should be able to make changes | 14:22 |
| *** yamamoto has joined #openstack-lbaas | 14:23 | |
| xgerman_ | periodic OSA | 14:26 |
| xgerman_ | translation scenario (which includes Octavia) | 14:26 |
| xgerman_ | http://paste.openstack.org/raw/619036/ | 14:28 |
| *** yamamoto has quit IRC | 14:28 | |
| xgerman_ | maybe some of the resources went away. Didn’t have yet had coffee so my thinking is slow ATM | 14:28 |
| voelzmo | @johnsom is there any place where you would collect ideas or a proposal about how to solve this before implementing a solution? I'd be interested in participating in the discussion if you start tackling that topic | 14:35 |
| xgerman_ | we were thinking using LB’s ACL’s to solve that issue | 14:36 |
| xgerman_ | there must be a spec floating around | 14:37 |
| *** aojea has quit IRC | 14:37 | |
| voelzmo | @xgerman_ LBs ACLs solve the issue to allow incoming traffic on loadbalanced members only from the LB? | 14:39 |
| voelzmo | I'd like to only expose what is needed on the members | 14:39 |
| *** aojea has joined #openstack-lbaas | 14:40 | |
| xgerman_ | well, that feels like a subnet design question | 14:43 |
| xgerman_ | unless you are worried about members talking to each other | 14:43 |
| voelzmo | I am | 14:45 |
| voelzmo | that's what I'm saying | 14:45 |
| voelzmo | I want to expose only what is necessary, even on my subnet | 14:45 |
| xgerman_ | ok, so you would need to put sec grps on members which allow our port | 14:46 |
| voelzmo | which works convenient on a lot of other IaaSes using some security group as marker for the LB | 14:46 |
| voelzmo | however, I just learned that this will no longer work, once octavia isn't a singleton anymore | 14:46 |
| xgerman_ | we haven’t crossed that bridge yet ;-) | 14:48 |
| xgerman_ | but I can see your issue I think we can solve that with a carefully chosen remote sec grp for you to allow | 14:48 |
| xgerman_ | then only LB traffic will hit your members | 14:49 |
| voelzmo | @xgerman_ sounds lovely, I'd like to stick to the concept of using marker security groups, if possible | 14:49 |
| xgerman_ | never heard of marker sec grps - so we might talk about the same thing | 14:50 |
| voelzmo | yeah, not sure what to call it | 14:50 |
| voelzmo | I'm using a security group just to reference it in another security group | 14:50 |
| voelzmo | to "mark" where traffic is coming from without having to specify IPs or network ranges | 14:51 |
| xgerman_ | yes, that’s a remote security group | 14:51 |
| voelzmo | thanks, that's what I'm talking about | 14:51 |
| xgerman_ | there is a similar concept in FWaaS V2 | 14:51 |
| xgerman_ | if you rather use firewalls in the future ;-) | 14:51 |
| voelzmo | @xgerman_ who knows about the future | 14:51 |
| xgerman_ | yes, indeed | 14:52 |
| *** belharar has quit IRC | 14:52 | |
| voelzmo | so even if I'm using a vendor-specific backend, I guess still something would need to be done in octavia to attach the same security group to all ports created, right? | 14:54 |
| voelzmo | so I could reference it as remote security group on my members | 14:54 |
| *** aojea has quit IRC | 14:54 | |
| *** aojea has joined #openstack-lbaas | 14:55 | |
| *** belharar has joined #openstack-lbaas | 14:56 | |
| xgerman_ | I would have to check if we don’t already do that and not expose but yes, we would likely have work ahead (but easy) | 14:56 |
| voelzmo | @xgerman_ I couldn't see it, at least | 14:57 |
| voelzmo | @johnsom the documentation still says you should create an attach your own security group to the created LB's port: https://docs.openstack.org/neutron/pike/admin/config-lbaas.html#building-an-lbaas-v2-load-balancer | 14:58 |
| voelzmo | you said this should only be possible for an admin? | 14:58 |
| xgerman_ | yes | 14:58 |
| voelzmo | I'm doing this as regular project member | 14:58 |
| xgerman_ | yeah, that also doesn’t sound right | 14:59 |
| xgerman_ | since we can have a different port than the VIP to reach out to the subnet | 14:59 |
| voelzmo | @xgerman_ exactly, that's one of my issues | 15:00 |
| xgerman_ | best would be to file. a bug and call it an RfE and we can go from there | 15:00 |
| johnsom | voelzmo Those are old neutron docs. Current docs are here: https://docs.openstack.org/octavia/latest/ | 15:00 |
| voelzmo | @johnsom thanks. "Old" still means they're part of pike, though, right? | 15:00 |
| xgerman_ | no | 15:00 |
| johnsom | Well, means someone added stuff to it that might not be accurate | 15:01 |
| johnsom | There are errors all over the neutron docs | 15:01 |
| voelzmo | @johnsom this has been in since mitaka and was at least accurate then | 15:01 |
| johnsom | We did a bunch of work in Pike to clean up and improve the docs | 15:02 |
| voelzmo | s/accurate/working/ | 15:02 |
| johnsom | I know there is some stuff in there that never was accurate | 15:02 |
| johnsom | voelzmo So path forward, it would be great to post a spec with a proposal so that folks can review and comment. | 15:03 |
| voelzmo | and by spec you mean an 'rfe' tagged bug in launchpad, or are you using a different process? | 15:04 |
| *** aojea has quit IRC | 15:05 | |
| johnsom | You could start as an RFE, but this seems like a big enough change that a spec might be best. | 15:05 |
| johnsom | https://github.com/openstack/octavia/tree/master/specs | 15:06 |
| johnsom | There is a template and example file there | 15:06 |
| *** cody-somerville has joined #openstack-lbaas | 15:06 | |
| *** cody-somerville has quit IRC | 15:06 | |
| *** cody-somerville has joined #openstack-lbaas | 15:06 | |
| voelzmo | okay, thanks | 15:06 |
| voelzmo | w00t, github, not gerrit? | 15:06 |
| voelzmo | No, I'm still proposing stuff on https://git.openstack.org, right? | 15:07 |
| xgerman_ | they are the same thing | 15:08 |
| voelzmo | ah, so much process confusion | 15:08 |
| voelzmo | well, github PRs and gerrit reviews are different things, right | 15:08 |
| *** aojea has joined #openstack-lbaas | 15:08 | |
| xgerman_ | yes, and yes | 15:09 |
| xgerman_ | gerrit review is what we aim for | 15:09 |
| voelzmo | fair enough, I'll try to come up with a spec and submit something for review | 15:09 |
| voelzmo | tbh, I know very little about the current internals of octavia, so this will be more of a wishlist, but we have something to talk about at least | 15:10 |
| voelzmo | thanks @johnsom and @xgerman_ | 15:10 |
| xgerman_ | especially how you liek that sec-grp be exposed to the user - API and CLI changes would be of interest. Slapping sec grp on ports is easy… | 15:10 |
| *** eezhova has quit IRC | 15:12 | |
| johnsom | Yeah, links to github are just the easiest way to point folks to files. It's still a standard OpenStack review process | 15:12 |
| johnsom | xgerman_ Do you have a link to the periodic job? | 15:13 |
| johnsom | "In the office now" and done with my first meeting. | 15:13 |
| *** rcernin has quit IRC | 15:15 | |
| *** belharar has quit IRC | 15:18 | |
| *** aojea has quit IRC | 15:22 | |
| *** yamamoto has joined #openstack-lbaas | 15:25 | |
| johnsom | xgerman_ rm_work To get Project Updates on the schedule, we need you to let us know if you (or a core team member who will be presenting in your place) are confirmed to attend Sydney—approval to go, a plane ticket, etc—by September 5. Project Updates in Sydney will be 20 minutes by default—we’ve found the 20 minute video recordings have better engagement than 40—but let us know if your update needs | 15:29 |
| johnsom | all 40 minutes. | 15:29 |
| *** yamamoto has quit IRC | 15:30 | |
| xgerman_ | ok | 15:34 |
| *** dougwig has joined #openstack-lbaas | 15:34 | |
| xgerman_ | johnsom I only have the link I posted earlier — I know the translation periodic job failed but no idea where to find it | 15:37 |
| johnsom | Ok, I think I found the job, looking now | 15:37 |
| xgerman_ | in the paste it looked like we couldn’t download some python stuff | 15:38 |
| xgerman_ | it’s stable/pike | 15:38 |
| johnsom | It looks like there were more infra mirror issues recently, so that might be the cause | 15:38 |
| xgerman_ | yep, it looked like something which should be intermittent | 15:39 |
| *** csomerville has joined #openstack-lbaas | 15:39 | |
| johnsom | xgerman_ So, if this is the job: http://status.openstack.org/openstack-health/#/job/periodic-openstack-ansible-deploy-translations-master-ubuntu-xenial | 15:42 |
| johnsom | It's failed once in the last seven days.... | 15:42 |
| johnsom | But the timestamps don't match up | 15:42 |
| *** cody-somerville has quit IRC | 15:43 | |
| *** belharar has joined #openstack-lbaas | 15:58 | |
| johnsom | Yeah, I don't see any octavia errors in there. The only error line in the paste is dpkg: error: error writing to '<standard output>': Broken pipe | 15:59 |
| johnsom | Which could just be some local issue or network issue. | 15:59 |
| johnsom | The periodic job (not the paste) was a infra issue IMO | 15:59 |
| xgerman_ | yeah, the paste was some guy running it on his box but that also had an error accessing some Internet resource | 16:02 |
| *** tesseract has quit IRC | 16:04 | |
| *** voelzmo has quit IRC | 16:06 | |
| *** eezhova has joined #openstack-lbaas | 16:10 | |
| *** catintheroof has joined #openstack-lbaas | 16:10 | |
| *** gcheresh has quit IRC | 16:22 | |
| *** belharar has quit IRC | 16:24 | |
| *** yamamoto has joined #openstack-lbaas | 16:26 | |
| *** yamamoto has quit IRC | 16:32 | |
| *** belharar has joined #openstack-lbaas | 16:33 | |
| *** pcaruana has quit IRC | 16:37 | |
| *** JudeC has joined #openstack-lbaas | 16:56 | |
| *** belharar has quit IRC | 17:00 | |
| *** gcheresh has joined #openstack-lbaas | 17:06 | |
| *** yamamoto has joined #openstack-lbaas | 17:28 | |
| *** yamamoto has quit IRC | 17:35 | |
| rm_work | hmmm | 17:46 |
| rm_work | that's the VIP Port | 17:46 |
| rm_work | that's the one we expose | 17:46 |
| rm_work | oh whoops scrolled up too much | 17:46 |
| *** tongl has joined #openstack-lbaas | 17:48 | |
| rm_work | johnsom: so does the core doing the project-update get a speaker pass? | 17:48 |
| rm_work | i have approval but i wouldn't have a badge at the moment | 17:48 |
| johnsom | You have the information I have, I can get you an e-mail to ask if you like | 17:53 |
| *** gcheresh has quit IRC | 17:56 | |
| *** JudeC has quit IRC | 18:01 | |
| *** eezhova has quit IRC | 18:04 | |
| nmagnezi | o/ | 18:04 |
| rm_work | hmm k | 18:09 |
| *** JudeC has joined #openstack-lbaas | 18:18 | |
| *** yamamoto has joined #openstack-lbaas | 18:31 | |
| *** voelzmo has joined #openstack-lbaas | 18:34 | |
| *** yamamoto has quit IRC | 18:37 | |
| *** gcheresh has joined #openstack-lbaas | 18:44 | |
| *** voelzmo has quit IRC | 18:46 | |
| *** amotoki has quit IRC | 18:53 | |
| xgerman_ | johnsom this combination of statistics and recent module is what I can comprehend pretty quickly: https://www.microsoft.com/developerblog/2017/06/28/scaling-udp-workloads-with-kubernetes/ | 19:02 |
| xgerman_ | downside is we probably need to rehash each time we add/remove LBs but we can fix that in the future with some of the better distributor stuff | 19:03 |
| xgerman_ | Then I can sling that on LB1 on an ACTIVE-PASSIVE set and they become magically ACTIVE-ACTIBE ;-) | 19:07 |
| *** yamamoto has joined #openstack-lbaas | 19:33 | |
| *** gcheresh has quit IRC | 19:35 | |
| *** gcheresh has joined #openstack-lbaas | 19:37 | |
| *** yamamoto has quit IRC | 19:38 | |
| *** eezhova has joined #openstack-lbaas | 19:44 | |
| *** armax has quit IRC | 19:52 | |
| *** fnaval has quit IRC | 19:56 | |
| *** sshank has joined #openstack-lbaas | 20:03 | |
| *** fnaval has joined #openstack-lbaas | 20:22 | |
| *** fnaval has quit IRC | 20:22 | |
| *** fnaval has joined #openstack-lbaas | 20:23 | |
| *** eezhova has quit IRC | 20:27 | |
| xgerman_ | rm_work we will get a feree code for attending the PTG | 20:33 |
| johnsom | Ah, true | 20:33 |
| xgerman_ | rm_work my approved budget only covers the flight and some meals… so if you know a nice bridge… | 20:34 |
| xgerman_ | I can probably kick in a few bucks if we want to go and do the update | 20:34 |
| *** yamamoto has joined #openstack-lbaas | 20:35 | |
| *** yamamoto has quit IRC | 20:38 | |
| *** yamamoto has joined #openstack-lbaas | 20:38 | |
| *** yamamoto has quit IRC | 20:39 | |
| *** rcernin has joined #openstack-lbaas | 20:50 | |
| *** gcheresh has quit IRC | 20:51 | |
| *** aojea has joined #openstack-lbaas | 21:02 | |
| *** fnaval has quit IRC | 21:04 | |
| *** armax has joined #openstack-lbaas | 21:04 | |
| *** yamamoto has joined #openstack-lbaas | 21:10 | |
| *** jmccrory has quit IRC | 21:13 | |
| *** jmccrory has joined #openstack-lbaas | 21:14 | |
| *** yamamoto has quit IRC | 21:15 | |
| *** kbyrne has quit IRC | 21:30 | |
| *** kbyrne has joined #openstack-lbaas | 21:32 | |
| rm_work | xgerman_: i might have room in an airbnb | 21:36 |
| rm_work | xgerman_: i can just get a 2-room one | 21:36 |
| rm_work | still way cheaper than hotel | 21:36 |
| rm_work | i am supposed to book that now, too | 21:36 |
| xgerman_ | awesome | 21:36 |
| rm_work | i've crashed on your couch a couple times I think :P | 21:37 |
| xgerman_ | yep | 21:37 |
| xgerman_ | which dates are you thiking? | 21:37 |
| rm_work | i'm still 100% verifying | 21:37 |
| xgerman_ | same here | 21:37 |
| rm_work | so i'll need till the end of the week prolly | 21:37 |
| xgerman_ | sounds good - | 21:46 |
| rm_work | johnsom / xgerman_: seems that keepalived is taking ~4s from when the MASTER dies to actually trigger transition to MASTER | 21:57 |
| rm_work | so I don't know how even upstream we are supposed to be doing subsecond | 21:57 |
| rm_work | because it's 4s from the last working request, until it does a GARP and theoretically would pick up new requests | 21:57 |
| johnsom | Oh, I never quoted sub-second, I usually say around a second | 21:57 |
| rm_work | lol | 21:57 |
| johnsom | Got back and watch the video from Tokyo | 21:58 |
| rm_work | well 4s seems a little sad | 21:58 |
| rm_work | and is not what I was expecting, quite | 21:58 |
| rm_work | i wonder if we can tweak that | 21:58 |
| rm_work | like, are any of our settings bad? | 21:58 |
| johnsom | Yeah, I don't think our defaults are "perfect" there | 21:58 |
| *** JudeC has quit IRC | 21:59 | |
| rm_work | hmm | 22:00 |
| johnsom | I think things have changed a bit too over time. | 22:00 |
| *** leitan has quit IRC | 22:01 | |
| *** catintheroof has quit IRC | 22:01 | |
| rm_work | oh nice, realized why storing topology on the LB is nice | 22:09 |
| rm_work | now I can switch the topo to ACTIVE_STANDBY and it won't do horrific things to the old LBs :P | 22:09 |
| rm_work | i'd never really thought about it | 22:09 |
| rm_work | anyway, about to switch over to active-standby... soon | 22:09 |
| *** yamamoto_ has joined #openstack-lbaas | 22:11 | |
| *** JudeC has joined #openstack-lbaas | 22:14 | |
| *** yamamoto_ has quit IRC | 22:17 | |
| rm_work | johnsom: wait is it related to: | 22:18 |
| rm_work | # vrrp_check_interval = 5 | 22:18 |
| rm_work | # vrpp_fail_count = 2 | 22:18 |
| rm_work | # vrrp_success_count = 2 | 22:18 |
| rm_work | that doesn't actually look like how long it takes | 22:18 |
| rm_work | I mean I guess that's just for the check script? | 22:18 |
| rm_work | unless it isn't? | 22:19 |
| *** sshank has quit IRC | 22:19 | |
| rm_work | yeah looks like those are used for the check script, not for the "pair liveness" | 22:20 |
| openstackgerrit | Adam Harwell proposed openstack/octavia master: WIP: Floating IP Network Driver (spans L3s) https://review.openstack.org/435612 | 22:21 |
| openstackgerrit | Adam Harwell proposed openstack/octavia master: Whitespace fixes and simplification of keepalived template https://review.openstack.org/496430 | 22:26 |
| rm_work | johnsom: ^^ that was bothering me a lot | 22:26 |
| rm_work | the template was an absolute mess | 22:26 |
| rm_work | the output had like 4 blank lines at the top, and EVERYTHING had different indentation | 22:26 |
| rm_work | it made me weep on the inside | 22:26 |
| johnsom | Hahaha | 22:27 |
| *** sshank has joined #openstack-lbaas | 22:27 | |
| rm_work | and like, why were we using a macro | 22:28 |
| rm_work | for one spot | 22:28 |
| rm_work | that was just a for-loop | 22:28 |
| rm_work | that's not a macro | 22:28 |
| rm_work | it still bothers me that there's two newlines at the end of the TEMPLATE, but in the output, there's only one for some freaking reason | 22:30 |
| rm_work | so whatever, it has to stay I guess >_< | 22:30 |
| *** rcernin has quit IRC | 22:32 | |
| rm_work | anyone know of any promo codes for the PTG? :P | 22:59 |
| *** aojea has quit IRC | 23:05 | |
| johnsom | Nope | 23:06 |
| *** ipsecguy has quit IRC | 23:07 | |
| *** ipsecguy has joined #openstack-lbaas | 23:08 | |
| tongl | If there is any exception raised during creating LB resource, how can we exit without any resource being created. In our vmware_nsx driver, in case of exception raised, it still go ahead to create LB in ERROR state in neutron db. | 23:11 |
| tongl | https://github.com/openstack/vmware-nsx/blob/master/vmware_nsx/services/lbaas/nsx_v3/loadbalancer_mgr.py#L43 | 23:11 |
| tongl | What I want is just output the exception to user, but not creating any resource in ERROR state in db. | 23:12 |
| johnsom | tongl Since the API is async you should always have a record for the client to query status | 23:12 |
| *** ipsecguy has quit IRC | 23:12 | |
| *** ipsecguy has joined #openstack-lbaas | 23:12 | |
| *** yamamoto_ has joined #openstack-lbaas | 23:13 | |
| tongl | If it fails validation at an early stage, is it possible to exit without creating any resource? | 23:13 |
| tongl | For me it makes sense since it didn't pass validation. It just exit without any resource being created. | 23:14 |
| johnsom | Yes, if it fails basic validation the user gets a direct response instead of 201, but once the data is validated it is stored in the DB for the driver to act on. | 23:15 |
| rm_work | tongl: yeah, by the time your driver gets it, it HAS passed validation | 23:16 |
| tongl | I see. So as long as the data passes basic validation, it will be created in db even if the driver validation part failed. | 23:16 |
| rm_work | yes, then it will go to ERROR and the user will know it failed | 23:17 |
| rm_work | though, what validation happens in your driver that is being allowed through the API validation layer? | 23:17 |
| rm_work | that's not ideal | 23:17 |
| tongl | Probably in our driver, we can delete the record if it failed the driver validation. | 23:17 |
| rm_work | tongl: err, that would be bad | 23:18 |
| rm_work | then the user would just see their LB disappear | 23:18 |
| rm_work | and be very confused | 23:18 |
| rm_work | they need to be able to query the LBID and see that it went to ERROR | 23:18 |
| johnsom | tongl That would be a bad idea. The point of 201 is to signal that this is an async interface and the client should poll for status | 23:18 |
| tongl | e.g $ neutron lbaas-loadbalancer-create --name lb1 private-subnet | 23:18 |
| *** yamamoto_ has quit IRC | 23:18 | |
| tongl | in our vmware-nsx driver, we requires there is a router for this subnet. otherwise it will fail lb validation. | 23:19 |
| rm_work | ok, so in that case, you would set it to ERROR | 23:19 |
| rm_work | and the user would be querying the LBID and see it go to ERROR status from PENDING_CREATE | 23:19 |
| tongl | so during lb creation, if it fails the subnet validation, we will raise BatRequest exception | 23:19 |
| rm_work | unfortunately for your case, the driver enters the picture too late to be part of API validation :/ | 23:20 |
| rm_work | it MIGHT be possible to allow for some custom validations... I've done a little bit of work on that front | 23:21 |
| tongl | I see. I thought it was because I didn't raise correct exception :) | 23:21 |
| rm_work | but that gets into murky territory | 23:21 |
| rm_work | no, by the time you even get the request, we've already accepted it and sent the user a 201 | 23:21 |
| johnsom | Yeah, by the time the driver gets the request the user is already gone | 23:22 |
| johnsom | This is something that could be proposed for the Octavia provider spec. Allowing drivers to have validator. | 23:23 |
| tongl | that would be awesome to allow driver validator | 23:23 |
| tongl | As I got some feedback that if it fails driver validator, it is confusing it the resource is still getting created in ERROR state. | 23:24 |
| johnsom | Yeah, it's an interesting case where drivers have additional restrictions | 23:25 |
| *** sshank has quit IRC | 23:43 | |
| *** tongl has quit IRC | 23:48 | |
Generated by irclog2html.py 2.15.3 by Marius Gedminas - find it at mg.pov.lt!
