| *** janki has quit IRC | 02:44 | |
| *** rh-jelabarre has quit IRC | 03:21 | |
| *** janki has joined #openstack-kuryr | 04:00 | |
| *** janki has quit IRC | 05:20 | |
| *** janki has joined #openstack-kuryr | 05:51 | |
| *** gcheresh has joined #openstack-kuryr | 06:05 | |
| *** pcaruana has joined #openstack-kuryr | 06:11 | |
| *** ccamposr has joined #openstack-kuryr | 07:08 | |
| *** aperevalov has joined #openstack-kuryr | 08:02 | |
| dulek | ltomasbo: I've debugged that ERROR happening in kuryr-daemon logs. Want a surprise? Same happens in n-cpu logs on gates with VM's. | 08:03 |
|---|---|---|
| dulek | ltomasbo: So it's more of OVN DevStack plugin bug. Who should I talk to about it? | 08:03 |
| ltomasbo | dulek, umm | 08:04 |
| ltomasbo | dulek, dalvarez most probably | 08:04 |
| dulek | ltomasbo: Great, thanks! | 08:04 |
| ltomasbo | dulek, have you tried to run the commands manually? | 08:05 |
| dulek | ltomasbo: Yes, they don't work. This isn't that complicated really, it's only because OVN's DevStack plugin puts OVN on a file socket. | 08:05 |
| ltomasbo | dulek, the command failing is similar to this one right: http://dani.foroselectronica.es/ovn-routing-and-ovn-trace-550/ | 08:05 |
| dulek | While the default from kuryr.conf/nova.conf is 127.0.0.1:6640. | 08:06 |
| ltomasbo | ahh, ok | 08:06 |
| dulek | ltomasbo: So it either needs to follow the default or [os_vif_ovs]ovsdb_i-don't-remember = <path to file socket> | 08:06 |
| dulek | or set. | 08:06 |
| ltomasbo | ahh, ok | 08:07 |
| dulek | Ah and not OVN on file socket, OVS DB is put on file socket. :P | 08:08 |
| * dulek is still drinking first coffee. | 08:08 | |
| ltomasbo | :) | 08:08 |
| *** openstackgerrit has quit IRC | 08:17 | |
| *** maysams has joined #openstack-kuryr | 08:21 | |
| *** gkadam has joined #openstack-kuryr | 08:33 | |
| *** gkadam has quit IRC | 08:33 | |
| *** alisanhaji has joined #openstack-kuryr | 08:50 | |
| *** celebdor has joined #openstack-kuryr | 09:07 | |
| *** pcaruana has quit IRC | 09:45 | |
| *** pcaruana has joined #openstack-kuryr | 09:46 | |
| *** alisanhaji has quit IRC | 10:40 | |
| dulek | maysams: Should I run the tests again? | 10:42 |
| maysams | dulek: That you be really great :) | 10:43 |
| dulek | maysams: Sure, no problem. | 10:43 |
| maysams | dulek: I was trying to set up everything manually | 10:43 |
| maysams | dulek: Thanks :D | 10:43 |
| maysams | s/you/would | 10:44 |
| *** alisanhaji has joined #openstack-kuryr | 10:54 | |
| *** openstackgerrit has joined #openstack-kuryr | 11:04 | |
| openstackgerrit | Maysa de Macedo Souza proposed openstack/kuryr-kubernetes master: Add support for text ports on Network Policy Spec https://review.openstack.org/643080 | 11:04 |
| openstackgerrit | Daniel Mellado proposed openstack/kuryr-kubernetes master: [WIP] Add ipBlock support to NP https://review.openstack.org/645139 | 11:36 |
| dmellado | maysams: ltomasbo that's just some idea that came to me about how to handle ipblock | 11:42 |
| dmellado | feel free to comment it there | 11:42 |
| ltomasbo | dmellado, checking! | 11:43 |
| ltomasbo | it may actually work! yes! I suppose you only need to cover the except case there | 11:45 |
| ltomasbo | dmellado, ^^ | 11:45 |
| dmellado | yep, I'll be adding that and a few test and check | 11:45 |
| ltomasbo | but yep, that was the direction I was thinking too (without paying attention to the corner cases...) | 11:45 |
| ltomasbo | dmellado, I guess we need to double check how that affects other stuff (services isolation, reaction to namespace, pod creation/re-labeling) | 11:46 |
| dmellado | yep, totally, in any case as a first glance they shouldn't be affected as it's just another set of cidrs to add | 11:46 |
| dmellado | which would only affect the ipblock cidr | 11:47 |
| dmellado | but let's see how the tests behave | 11:47 |
| dmellado | in any case ltomasbo maysams | 11:48 |
| dmellado | I saw that you did quite a few optimizations over my original code | 11:48 |
| dmellado | good work! xD | 11:48 |
| ltomasbo | yep, we hit a few corner cases... mostly about reacting to events | 11:48 |
| ltomasbo | pods creation/relabeling may not be affected | 11:49 |
| ltomasbo | namespaces may the same | 11:49 |
| ltomasbo | services.... could be more tricky... we need to double check | 11:49 |
| *** rh-jelabarre has joined #openstack-kuryr | 11:51 | |
| *** rh-jelabarre has quit IRC | 12:34 | |
| *** irclogbot_3 has quit IRC | 13:07 | |
| *** irclogbot_3 has joined #openstack-kuryr | 13:09 | |
| alisanhaji | Hi, I am trying to integrate Kuryr to Magnum, and I had a probelm with tiller deployment: http://paste.openstack.org/show/748168/ | 13:16 |
| alisanhaji | Kuryr kubernetes controller service didn't like the TargetPort as a string, it was expecting an int | 13:18 |
| alisanhaji | Should I file a bug report or was it already fixed? | 13:19 |
| dulek | alisanhaji: This should be fixed with https://review.openstack.org/#/c/641598/. | 13:23 |
| *** altlogbot_0 has quit IRC | 13:23 | |
| dulek | alisanhaji: Very recent thing. Do you use the container images from DockerHub? | 13:23 |
| *** altlogbot_1 has joined #openstack-kuryr | 13:24 | |
| alisanhaji | dulek: oh great, yes I do use the container images | 13:26 |
| alisanhaji | Very recent indeed :D | 13:27 |
| *** janki has quit IRC | 13:28 | |
| dulek | alisanhaji: I can try to update the DockerHub image for you, it'll just take a while, I have terrible Internet connection here. | 13:31 |
| *** gaoyan has joined #openstack-kuryr | 13:35 | |
| alisanhaji | That would be great, thanks! It's not urgent I am still having a misconfiguring LBaaS so I first need to deal with it, but it would be much easier for me if I could have the latest images to get everything working together | 13:35 |
| alisanhaji | dulek: misconfigured* | 13:36 |
| kmadac3 | Hi, I have deployed kuryr into VM with macvlan driver. I'm experiencing incorrect behavior. When I do simple deployment of 3 pods at one, those 3 pods are created, but IP/mac pair of only last created pod is added to allowed_address_pairs of neutron port. It seems like a bug to me. I checked the controller code, added couple of log messages and it seems that all 3 calls to get info about port to | 13:37 |
| kmadac3 | neutron are done at almost same time and allowed_address_pairs is empty at that moment, so only last update adds ip/mac to it. | 13:37 |
| *** altlogbot_1 has quit IRC | 13:39 | |
| *** altlogbot_3 has joined #openstack-kuryr | 13:40 | |
| dulek | kmadac3: Ah, seems like a race condition, isn't it? | 13:41 |
| kmadac3 | @dulek: is it seems so | 13:41 |
| dulek | kmadac3: We could use etags to fight it *if* Neutron API is supporting them. | 13:41 |
| kmadac3 | @dulek: should I create a bug for that? | 13:43 |
| dulek | kmadac3: Sure, but I doubt we have anyone who's going to grab this at the moment. | 13:43 |
| dulek | kmadac3: Wait, but shouldn't this happen for ip vlan case as well? | 13:44 |
| kmadac3 | @dulek: ok, I have some python development background so I can try to fix it | 13:44 |
| *** irclogbot_3 has quit IRC | 13:45 | |
| dulek | kmadac3: That's the fastest way IMO. I'm happy to help here. Let me check if Neutron API has etags. | 13:45 |
| kmadac3 | @dulek: great, I found something here: https://specs.openstack.org/openstack/api-wg/guidelines/etags.html | 13:46 |
| *** irclogbot_2 has joined #openstack-kuryr | 13:46 | |
| dulek | kmadac3: Yes, it's exactly this stuff and it seems Neutron does support it. | 13:46 |
| kmadac3 | @dulek: perfect, as etags are to me, i will need to study a bit, but I will do my best to implement it | 13:49 |
| dulek | kmadac3: Neutron's implementation may differ a bit as I look at it, but maybe I'm wrong. | 13:49 |
| kmadac3 | @dulek: * as etags are new to me | 13:49 |
| dulek | kmadac3: I don't have a cloud ready to check how it works at the moment, but will try this evening. | 13:50 |
| kmadac3 | @dulek: no problem, take your time | 13:50 |
| *** rh-jelabarre has joined #openstack-kuryr | 13:50 | |
| dulek | kmadac3: So Kuryr's implementation would just need to make sure to send If-Match along with this request: https://github.com/openstack/kuryr-kubernetes/blob/88e38e8e9493757d5db4da2b69dcc033b2f7465b/kuryr_kubernetes/controller/drivers/nested_macvlan_vif.py#L143-L146 | 13:52 |
| *** jistr is now known as jistr|call | 13:52 | |
| dulek | kmadac3: And also make sure it is retried a few times. | 13:52 |
| kmadac3 | @dulek: thanks, I will try to implement it there | 13:54 |
| *** jistr|call is now known as jistr | 14:00 | |
| maysams | dulek: ping | 14:05 |
| maysams | dulek, did you have a chance to run the tests again with my ps? | 14:05 |
| dulek | maysams: Yes. And no. :P | 14:15 |
| dulek | maysams: I have some troubles with etcd on my env and tests are failing due to that. | 14:15 |
| dulek | maysams: Will need to recreate it, but will only be able to do that later in the evening or tomorrow morning. | 14:16 |
| maysams | dulek, ahh okay | 14:16 |
| maysams | no worries, ty! | 14:16 |
| dulek | alisanhaji: The new kuryr-controller is pushed, it should include ltomasbo's fix. | 14:16 |
| ltomasbo | dulek, what fix? | 14:21 |
| dulek | ltomasbo: named ports in services. | 14:24 |
| ltomasbo | ahh, ok! | 14:24 |
| ltomasbo | so you have updated the upstream images? | 14:24 |
| *** gaoyan has quit IRC | 14:25 | |
| alisanhaji | dulek: great! Just as I was going to pull them, thanks | 14:31 |
| openstackgerrit | Daniel Mellado proposed openstack/kuryr-kubernetes master: [WIP] Add ipBlock support to NP https://review.openstack.org/645139 | 14:31 |
| kmadac3 | @dulek: I checked the response headers in 'openstack port show' output and there is no etag response header. Does it mean, that neutron doesnt support etag? I have Rocky release deployed and here the output: http://paste.openstack.org/show/748172/ | 14:32 |
| *** altlogbot_3 has quit IRC | 14:35 | |
| *** altlogbot_3 has joined #openstack-kuryr | 14:36 | |
| *** irclogbot_2 has quit IRC | 14:38 | |
| *** irclogbot_0 has joined #openstack-kuryr | 14:39 | |
| celebdor | kmadac3: etag? | 14:46 |
| celebdor | I never heard of etags :P | 14:47 |
| *** gcheresh has quit IRC | 14:50 | |
| *** gaoyan has joined #openstack-kuryr | 14:50 | |
| *** gaoyan has quit IRC | 14:51 | |
| dulek | ltomasbo: Yes. | 15:13 |
| dulek | alisanhaji: No problem. :) | 15:13 |
| dulek | kmadac3: I'll check in a moment, my env is up now. :) | 15:13 |
| dulek | celebdor: We have some lost update issue on mcvlan nested driver. | 15:14 |
| dulek | celebdor: And etags could help. It's something like resourceVersion on K8s. | 15:14 |
| dulek | maysams: http://paste.openstack.org/raw/748180/ | 15:29 |
| dulek | maysams: I'll post kuryr-controller log in a second. | 15:29 |
| maysams | dulek, okay | 15:33 |
| dulek | maysams: I've mailed it to you, sorry, didn't found a proper way to present a 2MB file. :P | 15:39 |
| dulek | maysams: Use the timestamp from the test to navigate - failure happened before 15:21:04.844. | 15:39 |
| dulek | (it may be way before) | 15:39 |
| dulek | (like ~5-10 minutes) | 15:40 |
| alisanhaji | dulek: Now everything is running, I can get magnum k8s to use kuryr, but when I schedule a pod, I have this error: NoMatches: No 'kuryr_kubernetes.vif_translators' driver found, looking for u'binding_failed' | 15:40 |
| maysams | dulek: Thanks a lot! | 15:40 |
| dulek | alisanhaji: Ah right, I saw that one day… | 15:40 |
| dulek | alisanhaji: This is in which Kuryr pod? | 15:41 |
| alisanhaji | When I look into neutron port, I find that the kuryr ones are DOWN and they have a binding_vif_type as binding_failed | 15:41 |
| alisanhaji | dulek: in the controller | 15:42 |
| dulek | alisanhaji: Okay, I think you should look into neutron logs. | 15:42 |
| dulek | alisanhaji: And I'm not sure which one. :P | 15:42 |
| dulek | alisanhaji: But it's probably port creation failure, IIRC last time someone asked about such error it was Neutron fautl. | 15:43 |
| dulek | fault | 15:43 |
| alisanhaji | the server logs only tell me that it failed to bind the port | 15:44 |
| alisanhaji | the neutron-agent get the RPC message for the new resource Port, but doesn't do anything | 15:44 |
| maysams | dulek: What sg drivers do you have enabled on your env? | 15:46 |
| dulek | maysams: Uh, oh. Maybe it's misconfigured? | 15:46 |
| dulek | maysams: pod_security_groups_driver = policy | 15:47 |
| dulek | It seems okay? | 15:47 |
| maysams | yup | 15:47 |
| dulek | alisanhaji: Sorry, I don't know Neutron well enough to help here. :( | 15:48 |
| maysams | dulek, and service_security_groups_driver ? | 15:48 |
| dulek | alisanhaji: But when I googled neutron binding_failed it seems to have some ideas. | 15:48 |
| dulek | maysams: Same. | 15:49 |
| maysams | okay, thanks | 15:49 |
| dulek | maysams: http://paste.openstack.org/raw/748183/ | 15:49 |
| maysams | dulek, that's correct | 15:50 |
| maysams | ty | 15:50 |
| alisanhaji | dulek: No worries, I will try to debug and see why it fails, thanks | 15:51 |
| dulek | maysams: Ah, one thing. If you'll see something very strange, like an K8s event missing from kuryr-controller logs, don't assume you're wrong. | 15:59 |
| dulek | maysams: I've had those issues with etcd, so maybe they're still manifesting. | 15:59 |
| dulek | And lost events were a sign of that on the gate. | 15:59 |
| maysams | dulek: I see lots of issues regarding missing annotation on namespace | 16:00 |
| maysams | dulek: Is that the one you were talking about? | 16:01 |
| dulek | maysams: No, no, I mean etcd failures may lead to some events not being passed to kuryr-controller's watcher. | 16:01 |
| maysams | dulek, right | 16:01 |
| dulek | maysams: But I've made some precautions to make sure etcd is fine, so it shouldn't happen. Just want to make sure you won't spend a day debugging, because you rule it out. ;) | 16:02 |
| dulek | maysams: And those missing annotations on namespace are because tests create a new namespace. | 16:02 |
| maysams | dulek: hahah ok | 16:02 |
| dulek | And do everything there. | 16:02 |
| maysams | yup yup | 16:02 |
| dulek | So obviously kuryr-controller needs some time to annotate it with the subnet created for it. | 16:02 |
| *** premsankar has joined #openstack-kuryr | 16:03 | |
| dulek | kmadac3: Okay, I've consulted that with Neutron folks: https://developer.openstack.org/api-ref/network/v2/#revisions | 16:08 |
| dulek | kmadac3: They don't have etags, they have revision_number. | 16:09 |
| dulek | kmadac3: And when doing PUT you can add "If-Match: <number>" to headers. | 16:09 |
| dulek | And request will fail if revision was changed. | 16:09 |
| dulek | kmadac3: It's even easier: def update_port(self, port, body=None, revision_number=None) | 16:11 |
| dulek | So update_port from here: https://github.com/openstack/kuryr-kubernetes/blob/88e38e8e9493757d5db4da2b69dcc033b2f7465b/kuryr_kubernetes/controller/drivers/nested_macvlan_vif.py#L143-L146 | 16:11 |
| dulek | Will accept revision_number parameter :) | 16:11 |
| maysams | dulek: Did deleted some security groups? I saw that the controller was over quota and then it got ready again | 16:14 |
| dulek | ltomasbo: This solves networking-ovn tracebacks: https://review.openstack.org/#/c/645096/ | 16:14 |
| dulek | maysams: I bumped up the quota number, I was building from your patch and it haven't included my fix for that. | 16:15 |
| dulek | maysams: So I did that manually after the cluster was up/. | 16:15 |
| maysams | dulek, right | 16:15 |
| dulek | maysams: I can run that test again, kill it while it's waiting for that pod and get the SG's and rules. | 16:16 |
| maysams | dulek, that would be great.. I'm thinking if the namespace used in the tests was created before you bumped the quota | 16:17 |
| dulek | maysams: No, it definitely was not. | 16:18 |
| maysams | ok | 16:18 |
| dulek | maysams: Okay, let's see… | 16:32 |
| dulek | maysams: Server is 10.1.1.242. client-a that cannot connect, but should is 10.1.1.254 | 16:32 |
| dulek | I have sg-network-policy-965-allow-client-a-via-named-port-egress-rule. | 16:32 |
| dulek | maysams: And those are the rules: http://paste.openstack.org/raw/748186/ | 16:33 |
| dulek | maysams: SG rule 0f1cce87-1673-4d24-b594-d84c17f83510 group is egress… | 16:34 |
| maysams | right, seems that it's creating the wrong rule | 16:36 |
| dulek | maysams: Nah, it doesn't look too bad yet IMO. | 16:37 |
| maysams | yet? | 16:37 |
| dulek | maysams: So LB VIP port has this attached: http://paste.openstack.org/raw/748188/ | 16:37 |
| dulek | Yet as I'm going through it. ;) | 16:37 |
| dulek | maysams: If I read that SG correctly it means that it allows ingress access from everywhere at port 80 and 81, right? | 16:38 |
| maysams | dulek, yup | 16:38 |
| dulek | And that seems fine. | 16:38 |
| dulek | Now the client-a pod… | 16:38 |
| dulek | maysams: http://paste.openstack.org/raw/748190/ | 16:39 |
| dulek | maysams: This is on client-a port. | 16:39 |
| kmadac3 | @dulek: sorry i was afk for a while. thanks for making things clear to me :) Hopefully it'll be enough for me to fix the issue. | 16:40 |
| dulek | maysams: That egress looks okay as well, isn't it? | 16:41 |
| maysams | dulek: yes, it does | 16:42 |
| dulek | kmadac3: It should be. Just pass port['revision_number'] to _update_port_address_pairs, add it to neutron.update_port and make sure it'll retry on 412 status code. | 16:43 |
| dulek | kmadac3: Where retry = fetching the port again and doing this: https://github.com/openstack/kuryr-kubernetes/blob/88e38e8e9493757d5db4da2b69dcc033b2f7465b/kuryr_kubernetes/controller/drivers/nested_macvlan_vif.py#L97-L113 | 16:44 |
| dulek | maysams: My only idea is that remote_ip_prefix='10.1.1.242/32' doesn't work as we thing? | 16:44 |
| dulek | s/thing/think | 16:44 |
| dulek | Maybe 10.1.1.242 doesn't fall into it, because it's not usable address. | 16:45 |
| dulek | But that's too bold, I wouldn't bet a cent on it… | 16:45 |
| maysams | dulek: how it would not be a usable address? | 16:46 |
| dulek | maysams: Good point, with /32 it is usable. :P | 16:46 |
| maysams | dulek, yup | 16:48 |
| maysams | dulek: could you check if the subnet got created for the namespace being used? | 16:57 |
| dulek | maysams: Sure, just a second. | 16:57 |
| dulek | maysams: cfdb78ed-a727-414b-b876-3f6c0fcdb098 | ns/network-policy-965-subnet | 4345f0b0-257f-47fb-8e9b-0da559c9fb52 | 10.1.1.192/26 | 16:57 |
| dulek | Looks pretty okay? | 16:57 |
| ltomasbo | dulek, nice! | 16:58 |
| ltomasbo | (the ovn fix) | 16:58 |
| dulek | maysams: I think that the server pod port doesn't allow any ingress. | 16:58 |
| dulek | maysams: http://paste.openstack.org/raw/748192/ | 16:59 |
| dulek | maysams: server pod's port has those SG's. | 16:59 |
| dulek | maysams: I don't see anything that would allow traffic from client-a or the LB? | 17:00 |
| alisanhaji | dulek: when kubelet tries to run kuryr-cni, I get an error: /opt/cni/bin/kuryr-cni: line 40: docker: command not found | 17:02 |
| maysams | dulek: me neither | 17:02 |
| maysams | ltomasbo: May you confirm if we only create a rule with remote_group_ig when namespace sgs driver is enabled: | 17:02 |
| dulek | alisanhaji: Hm, no docker on host with kubelet? Why? cri-o? | 17:02 |
| dulek | alisanhaji: Nah, you wouldn't see that in kubelet log if you had cri-o. So why no docker? :D | 17:03 |
| alisanhaji | dulek: Docker is present, but maybe the PATH is not present during kuryr-cni script | 17:04 |
| dulek | alisanhaji: Ah, might be, but I'm not sure what I can do here. We do `docker exec` to execute kuryr-cni inside our kuryr-cni container, so we have all the configs, etc. | 17:05 |
| dulek | maysams: Sorry, I don't understand. I don't want to reconfigure the env, it seems a bit fragile. ;) | 17:06 |
| dulek | maysams: I can try creating the SG rule myself and see if it helps. | 17:06 |
| maysams | dulek: No worries.. I will analyze better those SG you shared with me | 17:07 |
| maysams | thanks | 17:07 |
| alisanhaji | dulek: if you are running docker inside docker, I think I will need to mount /usr/bin with cni DaemonSet | 17:07 |
| dulek | maysams: This helped: openstack security group rule create --ingress --protocol tcp --remote-ip 10.1.1.233 1b76ff78-52df-42da-88c8-020c4999a97e | 17:09 |
| dulek | alisanhaji: It's not inside docker, it's run by kubelet. | 17:09 |
| dulek | Oh waaaait, your kubelet is inside docker container? | 17:09 |
| dulek | Nah, that shouldn't be, right? :P | 17:09 |
| dulek | alisanhaji: So if kuryr-cni script is run by kubelet we assume it's run on the host. | 17:10 |
| dulek | And by kuryr-cni I mean /opt/cni/bin/kuryr-cni | 17:10 |
| dulek | alisanhaji: Now tricky part is that this script only finds the kuryr-cni container through docker/runc and does docker/runc exec. | 17:10 |
| alisanhaji | no kubelet is not running inside a VM | 17:11 |
| *** maysams has quit IRC | 17:11 | |
| dulek | alisanhaji: So where is it running? | 17:12 |
| alisanhaji | Oh, it's actually running through runc | 17:14 |
| dulek | alisanhaji: Ah, so that error was a false positive? | 17:14 |
| dulek | alisanhaji: I really want to help here, but I need to go now. You can bug dmellado and ltomasbo about it if they're still online. Or maybe even celebdor if he still reads IRC. ;) | 17:15 |
| dulek | Guys, this is about Kuryr integration with Magnum, so pretty cool stuff in terms of K8s on OpenStack. | 17:15 |
| dmellado | hi o/ | 17:16 |
| dmellado | I was about to call it a day too, what's on? ;) | 17:16 |
| dulek | dmellado: If it's the case I don't keep you, go enjoy playing with the kid. ;) | 17:18 |
| alisanhaji | dulek: well I still have the error of not finding docker | 17:18 |
| alisanhaji | it occurs over and over | 17:18 |
| dulek | alisanhaji: Can you put full log somewhere? | 17:18 |
| dulek | alisanhaji: I mean full log of running kuryr-cni, not full kubelet log. | 17:19 |
| dmellado | so what's the issue? kuryr-cni is not running? | 17:19 |
| dmellado | on which context? | 17:19 |
| dulek | dmellado: For some reason on Magnum it keeps bugging alisanhaji with docker binary being not found. | 17:19 |
| dulek | That's the script by the way: https://github.com/openstack/kuryr-kubernetes/blob/master/cni_ds_init#L12-L53 | 17:20 |
| alisanhaji | here: http://paste.openstack.org/show/748193/ | 17:20 |
| alisanhaji | dulek: thanks | 17:20 |
| alisanhaji | why does it need to run kuryr-cni another time? | 17:21 |
| dulek | alisanhaji: YOu mean through `docker exec`? Because the main kuryr-cni is in Python and Python on the host isn't granted, also we don't have kuryr.conf there, etc. | 17:24 |
| dulek | It's easiest way, unfortunately Python apps aren't binaries that you can just put on host. | 17:24 |
| dulek | alisanhaji: Okay, so it seems like this fails: https://github.com/openstack/kuryr-kubernetes/blob/master/cni_ds_init#L41 | 17:25 |
| dulek | alisanhaji: And script falls back to use Docker. | 17:25 |
| dulek | And doesn't find it for some reason. | 17:25 |
| alisanhaji | dulek: I see | 17:25 |
| dulek | alisanhaji: Can't really help here, you need to figure out why kubelet processes doesn't see docker. And if there's a valid reason behind it we can discuss how to stop depending on it. | 17:26 |
| dulek | Okay, really need to go now. ;) | 17:26 |
| dulek | Have a great evening! | 17:26 |
| alisanhaji | dulek: ok I will try to, thanks for the help | 17:28 |
| *** alisanhaji has quit IRC | 17:30 | |
| *** gcheresh has joined #openstack-kuryr | 17:32 | |
| *** ccamposr has quit IRC | 17:36 | |
| *** gmann is now known as gmann_afk | 17:43 | |
| *** gcheresh has quit IRC | 17:58 | |
| *** maysams has joined #openstack-kuryr | 18:02 | |
| openstackgerrit | Daniel Mellado proposed openstack/kuryr-kubernetes master: [WIP] Add ipBlock support to NP https://review.openstack.org/645139 | 18:37 |
| *** gmann_afk is now known as gmann | 18:40 | |
| *** premsankar has quit IRC | 19:02 | |
| *** mrostecki has quit IRC | 19:35 | |
| *** mrostecki has joined #openstack-kuryr | 19:42 | |
| *** mrostecki has quit IRC | 19:43 | |
| *** mrostecki has joined #openstack-kuryr | 19:45 | |
| *** celebdor has quit IRC | 20:19 | |
| *** rh-jelabarre has quit IRC | 21:04 | |
| *** pcaruana has quit IRC | 21:33 | |
| *** celebdor has joined #openstack-kuryr | 21:33 | |
| *** maysams has quit IRC | 21:38 | |
| *** irclogbot_0 has quit IRC | 22:05 | |
Generated by irclog2html.py 2.15.3 by Marius Gedminas - find it at mg.pov.lt!