| *** neiljerram has quit IRC | 00:25 | |
| *** salv-orl_ has quit IRC | 00:27 | |
| *** limao has joined #openstack-kuryr | 00:58 | |
| openstackgerrit | Merged openstack/fuxi: Raise exception when find more than one matched Cinder volumes https://review.openstack.org/387937 | 01:24 |
|---|---|---|
| *** salv-orlando has joined #openstack-kuryr | 01:28 | |
| *** tonanhngo has joined #openstack-kuryr | 01:28 | |
| *** tonanhngo has quit IRC | 01:30 | |
| *** salv-orlando has quit IRC | 01:35 | |
| *** hongbin has joined #openstack-kuryr | 02:17 | |
| *** salv-orlando has joined #openstack-kuryr | 02:31 | |
| *** salv-orlando has quit IRC | 02:36 | |
| *** tonanhngo has joined #openstack-kuryr | 03:04 | |
| *** tonanhngo has quit IRC | 03:05 | |
| *** severion has joined #openstack-kuryr | 03:10 | |
| *** salv-orlando has joined #openstack-kuryr | 03:32 | |
| *** salv-orlando has quit IRC | 03:36 | |
| *** tonanhngo has joined #openstack-kuryr | 03:44 | |
| *** tonanhngo has quit IRC | 03:45 | |
| *** vikasc has quit IRC | 03:45 | |
| *** janki has joined #openstack-kuryr | 03:51 | |
| *** vikasc has joined #openstack-kuryr | 03:57 | |
| *** vikas_ has joined #openstack-kuryr | 04:01 | |
| *** vikasc has quit IRC | 04:03 | |
| *** hongbin has quit IRC | 04:07 | |
| *** hongbin has joined #openstack-kuryr | 04:07 | |
| openstackgerrit | Matt McEuen proposed openstack/kuryr-libnetwork: Add --gateway to README network create examples https://review.openstack.org/420493 | 04:09 |
| *** hongbin has quit IRC | 04:13 | |
| *** tonanhngo has joined #openstack-kuryr | 04:30 | |
| *** tonanhngo has quit IRC | 04:31 | |
| *** salv-orlando has joined #openstack-kuryr | 04:33 | |
| *** salv-orlando has quit IRC | 04:38 | |
| *** tonanhngo has joined #openstack-kuryr | 05:04 | |
| *** tonanhngo has quit IRC | 05:05 | |
| *** severion has quit IRC | 05:18 | |
| *** v1k0d3n has quit IRC | 05:18 | |
| *** salv-orlando has joined #openstack-kuryr | 05:34 | |
| *** v1k0d3n has joined #openstack-kuryr | 05:35 | |
| *** salv-orlando has quit IRC | 05:38 | |
| *** tonanhngo has joined #openstack-kuryr | 05:44 | |
| *** tonanhngo has quit IRC | 05:46 | |
| *** v1k0d3n has quit IRC | 05:48 | |
| *** v1k0d3n has joined #openstack-kuryr | 05:49 | |
| *** tonanhngo has joined #openstack-kuryr | 06:04 | |
| *** tonanhngo has quit IRC | 06:05 | |
| *** tonanhngo has joined #openstack-kuryr | 06:22 | |
| *** tonanhngo has quit IRC | 06:23 | |
| *** pmannidi has quit IRC | 06:29 | |
| *** pmannidi has joined #openstack-kuryr | 06:32 | |
| *** pmannidi is now known as pmannidi|mtg | 06:32 | |
| *** salv-orlando has joined #openstack-kuryr | 06:35 | |
| *** salv-orlando has quit IRC | 06:39 | |
| *** pmannidi|mtg is now known as pmannidi | 07:00 | |
| *** yamamoto has quit IRC | 07:04 | |
| *** pmannidi has quit IRC | 07:04 | |
| *** pmannidi has joined #openstack-kuryr | 07:18 | |
| *** vikas_ has quit IRC | 07:25 | |
| *** tonanhngo has joined #openstack-kuryr | 07:27 | |
| *** tonanhngo has quit IRC | 07:29 | |
| *** vikas_ has joined #openstack-kuryr | 07:29 | |
| *** pcaruana has joined #openstack-kuryr | 07:34 | |
| *** salv-orlando has joined #openstack-kuryr | 07:36 | |
| *** salv-orlando has quit IRC | 07:40 | |
| *** roeyc has joined #openstack-kuryr | 07:51 | |
| *** tonanhngo has joined #openstack-kuryr | 08:21 | |
| *** tonanhngo has quit IRC | 08:22 | |
| *** roeyc has quit IRC | 08:26 | |
| *** salv-orlando has joined #openstack-kuryr | 08:37 | |
| *** salv-orlando has quit IRC | 08:41 | |
| *** tonanhngo has joined #openstack-kuryr | 08:46 | |
| *** tonanhngo has quit IRC | 08:47 | |
| openstackgerrit | Dongcan Ye proposed openstack/kuryr-libnetwork: Removes subnetpool_id tag for Neutron existing subnet https://review.openstack.org/419735 | 09:00 |
| *** roeyc has joined #openstack-kuryr | 09:29 | |
| *** garyloug has joined #openstack-kuryr | 09:33 | |
| *** vikas_ has quit IRC | 09:35 | |
| *** salv-orlando has joined #openstack-kuryr | 09:37 | |
| *** yedongcan has joined #openstack-kuryr | 09:39 | |
| *** salv-orlando has quit IRC | 09:42 | |
| *** saneax-_-|AFK is now known as saneax | 09:43 | |
| *** vikas_ has joined #openstack-kuryr | 09:53 | |
| *** roeyc has quit IRC | 09:59 | |
| *** roeyc has joined #openstack-kuryr | 10:01 | |
| *** limao has quit IRC | 10:07 | |
| *** devvesa has joined #openstack-kuryr | 10:09 | |
| *** roeyc has quit IRC | 10:14 | |
| *** yedongcan has left #openstack-kuryr | 10:38 | |
| *** salv-orlando has joined #openstack-kuryr | 10:38 | |
| *** salv-orlando has quit IRC | 10:43 | |
| *** neiljerram has joined #openstack-kuryr | 10:51 | |
| janonymous | limao , ivc_ : Congrats! Well deserved cores :) | 11:00 |
| apuimedo | indeed! | 11:05 |
| janonymous | :) | 11:06 |
| *** tonanhngo has joined #openstack-kuryr | 11:06 | |
| *** tonanhngo has quit IRC | 11:08 | |
| *** salv-orlando has joined #openstack-kuryr | 11:39 | |
| ivc_ | janonymous apuimedo thanks :) | 11:40 |
| apuimedo | thanks to you for the hard work! | 11:40 |
| *** salv-orlando has quit IRC | 11:43 | |
| vikas_ | ivc_, congrats :) | 11:45 |
| *** roeyc has joined #openstack-kuryr | 11:49 | |
| *** saneax is now known as saneax-_-|AFK | 12:01 | |
| *** janki has quit IRC | 12:18 | |
| *** janki has joined #openstack-kuryr | 12:20 | |
| *** salv-orlando has joined #openstack-kuryr | 12:40 | |
| *** salv-orlando has quit IRC | 12:44 | |
| *** salv-orlando has joined #openstack-kuryr | 12:46 | |
| *** janki has quit IRC | 12:58 | |
| *** dougbtv has joined #openstack-kuryr | 13:00 | |
| *** janki has joined #openstack-kuryr | 13:06 | |
| *** garyloug has quit IRC | 13:34 | |
| *** janki has quit IRC | 13:38 | |
| *** salv-orlando has quit IRC | 13:46 | |
| *** limao has joined #openstack-kuryr | 13:49 | |
| *** yedongcan has joined #openstack-kuryr | 13:54 | |
| *** saneax-_-|AFK is now known as saneax | 13:59 | |
| *** limao has quit IRC | 14:00 | |
| *** garyloug has joined #openstack-kuryr | 14:00 | |
| *** limao has joined #openstack-kuryr | 14:02 | |
| *** tonanhngo has joined #openstack-kuryr | 14:08 | |
| *** tonanhngo has quit IRC | 14:10 | |
| *** mattmceuen has joined #openstack-kuryr | 14:11 | |
| *** janki has joined #openstack-kuryr | 14:16 | |
| openstackgerrit | Merged openstack/kuryr-kubernetes: docs: Fix image rendering https://review.openstack.org/419981 | 14:18 |
| *** hongbin has joined #openstack-kuryr | 14:27 | |
| *** janki has quit IRC | 14:41 | |
| *** v1k0d3n has quit IRC | 14:43 | |
| *** salv-orlando has joined #openstack-kuryr | 14:47 | |
| openstackgerrit | Merged openstack/kuryr-libnetwork: Add --gateway to README network create examples https://review.openstack.org/420493 | 14:50 |
| *** salv-orlando has quit IRC | 14:51 | |
| apuimedo | ivc_: My reading is that it behaves the same | 15:00 |
| apuimedo | being that we do not expose the containers to the hosts | 15:01 |
| ivc_ | apuimedo i think with k8s externalIP you can have multiple services sharing the same external IP | 15:01 |
| ivc_ | apuimedo but if we use floating ip to expose lbaas that would be different | 15:02 |
| *** yedongcan has left #openstack-kuryr | 15:03 | |
| ivc_ | apuimedo so imo in our case floating-ip actually maps to 'loadbalancer' type service and i'm not sure if we can support external IP | 15:05 |
| apuimedo | there's no question that floating-ip maps to loadbalancer type | 15:07 |
| *** hongbin has quit IRC | 15:08 | |
| ivc_ | apuimedo its just a bit confusing that we have lbaas for both clusterIP and loadbalancer type services | 15:09 |
| apuimedo | I think it's quite useful while we don't have loadbalancer type support (which IIRC requires implementing a cloud provider in K8s) | 15:09 |
| *** roeyc has quit IRC | 15:11 | |
| openstackgerrit | Merged openstack/fuxi: Make py35 devstack gate working https://review.openstack.org/419683 | 15:17 |
| mchiappero | sorry, back to the previous race topic, so, potentially batch of subsequent calls to neutron could be unsafe, right? | 15:22 |
| mchiappero | *any batch | 15:22 |
| *** saneax is now known as saneax-_-|AFK | 15:22 | |
| ltomasbo | yes, calls in OpenStack are mostly async | 15:27 |
| ltomasbo | so, as when you call trunk_add_subport it also internally calls update_port API, it happens that, as we also call update_port from kuryr-libnetwork | 15:29 |
| ltomasbo | the first part of trunk_add_subport gets executed first, then the kuryr-libnetwork update_port call, and finally the trunk_add_subport call to update_port | 15:29 |
| ivc_ | ltomasbo that's not exactly true. if you made an api request and got a response, the data in that response is already commited | 15:30 |
| ltomasbo | yes, if the call is sync, yes | 15:30 |
| mchiappero | no ok, I understood that specific problem, I'm wondering whether there are other part of the code potentially with that issue | 15:31 |
| ltomasbo | and when you perform the update_port and you received the response, then it is commited | 15:31 |
| mchiappero | is there a way to batch calls? | 15:31 |
| mchiappero | and of course execute them atomically | 15:31 |
| ltomasbo | the problem is that trunk_add_subports make a few tasks, and one of then is to call update_port | 15:32 |
| mchiappero | no ok, but what if I update the port and read back, do I have the guarantee the subsequent read will contain the updated info? | 15:32 |
| mchiappero | besides that specific trunk_add_subports problem | 15:33 |
| ltomasbo | when you update the port, the subsequent call will have the updated info | 15:33 |
| mchiappero | I mean in general | 15:33 |
| ltomasbo | the problem is that there is another call in between your write and your read | 15:33 |
| ltomasbo | changing the info | 15:33 |
| mchiappero | ? | 15:34 |
| ltomasbo | I mean, if 2 entities trigger two update ports at the same time, you cannot know which one will be the first one and the second one being executed | 15:34 |
| mchiappero | broadly speaking, what are the guarantees from neutron? | 15:34 |
| ltomasbo | so, it could lead to the wrong information (if there was a specific order) | 15:35 |
| mchiappero | ltomasbo: that makes perfectly sense | 15:35 |
| mchiappero | but what if I have an update and a read? | 15:35 |
| ltomasbo | then, the read should have the updated info (unless there is a cache method keeping the value for a certain time) | 15:36 |
| mchiappero | also, I'm thinking whether batching is supported | 15:36 |
| ltomasbo | not sure, but I think it goes against scalability design | 15:37 |
| mchiappero | I don't see how | 15:38 |
| mchiappero | anyway, still, I don't fully understand whether the return code from neutron means "enqueued" or "done" | 15:38 |
| mchiappero | because if it's the latter then the race shouldn't be there | 15:39 |
| mchiappero | (end if it's the former support for atomic changes/batches should be there) | 15:41 |
| *** tonanhngo has joined #openstack-kuryr | 15:41 | |
| *** tonanhngo has quit IRC | 15:42 | |
| ltomasbo | seems it is done in the sense it is written in the database | 15:43 |
| ltomasbo | but that for single calls, the problem here is that the trunk_subport_add is also calling update_port, and that is not ensured (maybe it is even a bug...) | 15:44 |
| ivc_ | ltomasbo do you mean that when you have 'update_port' (a) followed by 'trunk_subport_add' (b) and then 'show_port' (c) it is possible to get either 'kuryr:container' from (a) or 'trunk:subport' from (b) in (c) query? | 15:47 |
| *** limao has quit IRC | 15:47 | |
| ivc_ | ltomasbo or is it always 'trunk:subport'? | 15:47 |
| *** salv-orlando has joined #openstack-kuryr | 15:48 | |
| ltomasbo | in my case, the kuryr code does this: | 15:49 |
| ltomasbo | - trunk_subport_add | 15:49 |
| ltomasbo | - update_port | 15:49 |
| *** hongbin has joined #openstack-kuryr | 15:49 | |
| ltomasbo | and it ends up with turnk:subport | 15:49 |
| ltomasbo | if I do this: | 15:49 |
| ltomasbo | -trunk_subport_add | 15:49 |
| ltomasbo | - sleep 1 | 15:49 |
| ltomasbo | - update_port | 15:49 |
| ltomasbo | it ends up with kuryr:container | 15:49 |
| ivc_ | ltomasbo in the first case, is update_port called after successful return from 'trunk_subport_add' or are they both called async? | 15:51 |
| ltomasbo | just using the neutron_client, one after another in the kuryr-libnetwork code | 15:51 |
| *** salv-orlando has quit IRC | 15:52 | |
| ltomasbo | https://github.com/openstack/kuryr-libnetwork/blob/master/kuryr_libnetwork/port_driver/drivers/vlan.py#L61:L64 | 15:52 |
| ivc_ | got it. it seems that we need to dig into neutron trunk impl to find the part that does that async update. i wonder why is it async in the first place | 15:57 |
| apuimedo | probably a case of premature optimization | 15:57 |
| *** apuimedo is now known as apuimedo|away | 15:57 | |
| ivc_ | apuimedo|away ltomasbo it's worse | 16:00 |
| ivc_ | ltomasbo https://github.com/openstack/neutron/blob/master/neutron/services/trunk/rpc/server.py#L150-L166 | 16:02 |
| ivc_ | its triggered by agent | 16:02 |
| mchiappero | I haven't checked the code, but that's exactly my question... besides being handled asynchronously or not, what's the contract? What can I expect to be done in Neutron once trunk_subport_add (or whatever call) returns? | 16:08 |
| mchiappero | at least we know whether there is a bug by definition or other code is potentially broken in kuryr | 16:09 |
| ltomasbo | I found this: # NOTE(status_police) Set the trunk in BUILD state before processing | 16:10 |
| ltomasbo | # subport bindings. The trunk will stay in BUILD state until an | 16:10 |
| ltomasbo | # attempt has been made to bind all subports passed here and the | 16:10 |
| ltomasbo | # agent acknowledges the operation was successful. | 16:10 |
| mchiappero | in general, I would not expect such timing issues not to happen, or provide facilities to work with them, which is what I meant above. Then extending the API can be okay anyway, but still, I feel there is something broken in neutron | 16:10 |
| mchiappero | sorry, I s/I would not/I would/ | 16:11 |
| mchiappero | uff.. also remove "I" :D | 16:11 |
| ivc_ | ltomasbo we probably need a 'wait-loop' for trunk similar to what we have for port status=ACTIVE | 16:22 |
| ltomasbo | that is one of the solutions | 16:22 |
| ltomasbo | but it will delay container booting more | 16:22 |
| ivc_ | ltomasbo but we need to delay for the same reason as for port's status=ACTIVE | 16:23 |
| ivc_ | otherwise you could probably get a container that does not have network access | 16:23 |
| ltomasbo | why no network access? | 16:24 |
| ivc_ | because trunk is configured async | 16:24 |
| ivc_ | so if you just add a subport and fire up a container you might end up with a running container while neutron/agent are still configuring subport | 16:25 |
| ivc_ | ltomasbo btw how will it work if neutron agent is down? | 16:26 |
| ivc_ | will it start a container now? | 16:26 |
| mchiappero | I still don't understand why this is possible, every async calls provide a callback/notification system, or batch calls | 16:27 |
| ltomasbo | ivc_, I though you meant for the trunk_subport_add | 16:27 |
| ltomasbo | s/though/thought | 16:27 |
| ivc_ | ltomasbo yes i mean trunk_subport_add | 16:27 |
| ltomasbo | the wait for port status = ACTIVE is there anyway | 16:27 |
| ltomasbo | that is in a different kuryr-libnetwork/controller.py step, waiting for creating the container | 16:28 |
| ltomasbo | so, that part is exactly the same as for normal ports | 16:28 |
| ivc_ | ok but still 'trunk_subport_add' have pretty much the same async behaviour | 16:28 |
| *** tonanhngo has joined #openstack-kuryr | 16:28 | |
| ivc_ | mchiappero its not completely async. it does some work and returns a state with BUILD status and you can poll it for updates | 16:30 |
| ivc_ | mchiappero unfortunately the 'notification' mechanism is sort of lacking | 16:30 |
| *** salv-orlando has joined #openstack-kuryr | 16:30 | |
| mchiappero | ivc_: whatever... it's bad/incomplete design | 16:31 |
| mchiappero | which is worrisome | 16:31 |
| ivc_ | e.g. we have a nova notifier that for port's status=ACTIVE, but it's just for nova | 16:31 |
| mchiappero | so, let's poll then, since that's the expected approach :) | 16:32 |
| ivc_ | mchiappero thats what we do now for port's status in both kuryr-libnetwork and kuryr-k8s | 16:32 |
| mchiappero | ivc_: one more reason to do the same (I didn't know, I never noticed) | 16:33 |
| ivc_ | mchiappero tho at some point we could add another method using neutron 'notification driver' (meaning we'll connect to neutron's mq exchange) | 16:33 |
| *** tonanhngo has quit IRC | 16:34 | |
| mchiappero | ivc_: whatever works :) | 16:35 |
| ivc_ | ltomasbo mchiappero imo the right way to do it as ltomasbo has shown with "sleep(1)" between trunk and port update, except instead of 'sleep' we should poll for trunk status > BUILD | 16:35 |
| mchiappero | yeah ok | 16:36 |
| ivc_ | ltomasbo but ofc if we can get neutron team to update trunk_subport_add to accept device_owner, nothing of that would be necessary | 16:36 |
| mchiappero | yes, I'm not against that, as I said, that's okay anyway | 16:37 |
| mchiappero | but still, every single piece of software on earth needs to provide a well defined contract and provide a notification to the caller of the completion, either naturally by returning on blocking calls or by means of notifications/callbacks for non-blocking calls | 16:38 |
| mchiappero | it seems this is missing here | 16:38 |
| mchiappero | that's my only complain :) | 16:39 |
| ltomasbo | :D | 16:42 |
| ivc_ | mchiappero dreamer xD | 16:58 |
| *** devvesa has quit IRC | 17:09 | |
| mchiappero | naaa | 17:09 |
| mchiappero | but seriously, it's a pity kuryr has to poll for a deficiency present elsewhere | 17:10 |
| mchiappero | anyway... :) | 17:10 |
| mchiappero | LOL, we might start polling people from neutron to begin with :P | 17:11 |
| openstackgerrit | OpenStack Proposal Bot proposed openstack/fuxi: Updated from global requirements https://review.openstack.org/419929 | 17:18 |
| openstackgerrit | OpenStack Proposal Bot proposed openstack/kuryr: Updated from global requirements https://review.openstack.org/418792 | 17:20 |
| *** tonanhngo has joined #openstack-kuryr | 17:37 | |
| *** neiljerram has quit IRC | 17:41 | |
| *** tonanhngo has quit IRC | 17:42 | |
| *** roeyc has joined #openstack-kuryr | 17:49 | |
| *** roeyc has quit IRC | 17:49 | |
| *** garyloug has quit IRC | 18:07 | |
| *** tonanhngo has joined #openstack-kuryr | 18:10 | |
| *** tonanhngo has quit IRC | 18:11 | |
| *** portdirect is now known as intlabs | 18:14 | |
| *** intlabs is now known as portdirect | 18:17 | |
| *** tonanhngo has joined #openstack-kuryr | 18:33 | |
| *** salv-orlando has quit IRC | 18:35 | |
| *** salv-orlando has joined #openstack-kuryr | 18:36 | |
| *** tonanhngo has quit IRC | 18:38 | |
| *** tonanhngo has joined #openstack-kuryr | 18:49 | |
| *** tonanhngo has quit IRC | 18:49 | |
| *** tonanhngo has joined #openstack-kuryr | 18:59 | |
| *** tonanhngo has quit IRC | 18:59 | |
| *** tonanhngo has joined #openstack-kuryr | 19:19 | |
| *** tonanhngo has quit IRC | 19:20 | |
| *** tonanhngo has joined #openstack-kuryr | 19:29 | |
| *** tonanhngo has quit IRC | 19:30 | |
| *** salv-orlando has quit IRC | 20:23 | |
| openstackgerrit | OpenStack Proposal Bot proposed openstack/kuryr: Updated from global requirements https://review.openstack.org/418792 | 20:31 |
| openstackgerrit | OpenStack Proposal Bot proposed openstack/kuryr-libnetwork: Updated from global requirements https://review.openstack.org/419932 | 20:31 |
| *** salv-orlando has joined #openstack-kuryr | 21:09 | |
| *** tonanhngo has joined #openstack-kuryr | 21:23 | |
| *** tonanhngo has quit IRC | 21:23 | |
| *** tonanhngo has joined #openstack-kuryr | 21:28 | |
| *** salv-orlando has quit IRC | 21:47 | |
| *** v1k0d3n has joined #openstack-kuryr | 21:50 | |
| *** salv-orlando has joined #openstack-kuryr | 22:12 | |
| *** yamamoto has joined #openstack-kuryr | 22:15 | |
| *** salv-orl_ has joined #openstack-kuryr | 22:24 | |
| *** salv-orlando has quit IRC | 22:26 | |
| *** v1k0d3n has quit IRC | 23:08 | |
| *** salv-orl_ has quit IRC | 23:50 | |
| *** mattmceuen has quit IRC | 23:57 | |
Generated by irclog2html.py 2.14.0 by Marius Gedminas - find it at mg.pov.lt!