| *** diga has joined #openstack-kuryr | 00:00 | |
| *** tonanhngo has joined #openstack-kuryr | 00:40 | |
| *** limao has joined #openstack-kuryr | 00:41 | |
| *** tonanhngo has quit IRC | 00:43 | |
| *** diga has quit IRC | 01:21 | |
| *** yuanying has joined #openstack-kuryr | 01:24 | |
| *** yuanying has quit IRC | 01:24 | |
| *** salv-orl_ has joined #openstack-kuryr | 02:29 | |
| *** salv-orlando has quit IRC | 02:32 | |
| *** limao has quit IRC | 02:41 | |
| *** tonanhngo has joined #openstack-kuryr | 02:42 | |
| *** limao has joined #openstack-kuryr | 02:43 | |
| *** tonanhngo has quit IRC | 02:44 | |
| *** limao has quit IRC | 02:46 | |
| *** limao has joined #openstack-kuryr | 02:46 | |
| *** hongbin_ has joined #openstack-kuryr | 02:50 | |
| *** yuanying has joined #openstack-kuryr | 02:52 | |
| *** yuanying has quit IRC | 02:55 | |
| *** vikasc has quit IRC | 02:56 | |
| *** yuanying has joined #openstack-kuryr | 02:57 | |
| *** vikasc has joined #openstack-kuryr | 02:57 | |
| openstackgerrit | Liping Mao proposed openstack/kuryr-libnetwork: [NOT MERGE] Test Jenkins https://review.openstack.org/365683 | 02:59 |
|---|---|---|
| *** yuanying has quit IRC | 03:00 | |
| *** yuanying has joined #openstack-kuryr | 03:00 | |
| *** tonanhngo has joined #openstack-kuryr | 03:02 | |
| *** tonanhngo has quit IRC | 03:04 | |
| openstackgerrit | Liping Mao proposed openstack/kuryr-libnetwork: Fix wrong dir in post_test_hook.sh https://review.openstack.org/365827 | 03:08 |
| *** janki has joined #openstack-kuryr | 03:09 | |
| *** yuanying has quit IRC | 03:18 | |
| *** yuanying has joined #openstack-kuryr | 03:20 | |
| *** tonanhngo has joined #openstack-kuryr | 03:23 | |
| *** tonanhngo has quit IRC | 03:24 | |
| *** diga has joined #openstack-kuryr | 03:51 | |
| *** janki has quit IRC | 03:51 | |
| *** tonanhngo has joined #openstack-kuryr | 03:52 | |
| *** tonanhngo has quit IRC | 03:54 | |
| *** yuanying has quit IRC | 04:00 | |
| *** hongbin_ has quit IRC | 04:01 | |
| *** yuanying has joined #openstack-kuryr | 04:02 | |
| limao | hello janonymous | 04:04 |
| *** sdake has joined #openstack-kuryr | 04:05 | |
| janonymous | limao : hey | 04:05 |
| limao | hi janonymous, https://review.openstack.org/#/c/355786/ | 04:06 |
| limao | about this patch, do you mind rebase this patch base on my patch https://review.openstack.org/365827 ? | 04:06 |
| janonymous | limao: No prob ! | 04:08 |
| limao | Cool janonymous, Let's see if it can fix fullstack problem | 04:09 |
| *** tonanhngo has joined #openstack-kuryr | 04:15 | |
| openstackgerrit | Liping Mao proposed openstack/kuryr-libnetwork: [NOT MERGE] Test Jenkins https://review.openstack.org/365683 | 04:16 |
| *** tonanhngo has quit IRC | 04:16 | |
| *** sdake has quit IRC | 04:25 | |
| openstackgerrit | Liping Mao proposed openstack/kuryr-libnetwork: Fix Rally test job in kuryr-libnetwork https://review.openstack.org/365851 | 04:26 |
| openstackgerrit | Liping Mao proposed openstack/kuryr-libnetwork: [WIP]Fix Rally test job in kuryr-libnetwork https://review.openstack.org/365851 | 04:32 |
| openstackgerrit | janonymous proposed openstack/kuryr-libnetwork: Modify fullstack tests https://review.openstack.org/365856 | 04:46 |
| *** sdake has joined #openstack-kuryr | 04:50 | |
| janonymous | limao: i rebased a new patch...forgot to mention change id..anyway i am abandoning my previous patch now... Please see patch 365856 | 04:51 |
| limao | Cool， thanks | 04:59 |
| *** sdake has quit IRC | 05:01 | |
| openstackgerrit | Liping Mao proposed openstack/kuryr-libnetwork: Fix Rally test job in kuryr-libnetwork https://review.openstack.org/365851 | 05:03 |
| *** tonanhngo has joined #openstack-kuryr | 05:03 | |
| *** irenab has joined #openstack-kuryr | 05:04 | |
| *** tonanhngo has quit IRC | 05:04 | |
| *** oanson has joined #openstack-kuryr | 05:05 | |
| limao | Hi janonymous, fullstack still error | 05:05 |
| limao | maybe it is because https://review.openstack.org/#/c/363414/ | 05:05 |
| *** irenab_ has joined #openstack-kuryr | 05:06 | |
| limao | 2016-09-06 04:59:18.066620 | 2016-09-06 04:59:18.066 | File "/opt/stack/new/kuryr-libnetwork/.tox/fullstack/local/lib/python2.7/site-packages/docker/client.py", line 145, in _raise_for_status | 05:07 |
| limao | 2016-09-06 04:59:18.070881 | 2016-09-06 04:59:18.070 | raise errors.NotFound(e, response, explanation=explanation) | 05:07 |
| limao | 2016-09-06 04:59:18.075089 | 2016-09-06 04:59:18.074 | docker.errors.NotFound: 404 Client Error: NOT FOUND ("{ | 05:07 |
| limao | 2016-09-06 04:59:18.079468 | 2016-09-06 04:59:18.079 | "Err": "404: Not Found" | 05:07 |
| limao | 2016-09-06 04:59:18.083636 | 2016-09-06 04:59:18.083 | }") | 05:07 |
| *** irenab has quit IRC | 05:08 | |
| *** irenab_ is now known as irenab | 05:08 | |
| janonymous | yes..it is related.. | 05:10 |
| *** sdake has joined #openstack-kuryr | 05:10 | |
| janonymous | limao: i will push a patch to add env var of DOCKER_HOST also which if not set gives such errors | 05:11 |
| janonymous | limao: but i am not 100% sure if it solves the errors, i am looking in it now | 05:15 |
| limao | Let me try janonymous in the [NOT MERGE] Test Jenkins | 05:16 |
| openstackgerrit | Liping Mao proposed openstack/kuryr-libnetwork: [NOT MERGE] Test Jenkins https://review.openstack.org/365683 | 05:16 |
| *** sdake has quit IRC | 05:17 | |
| limao | janonymous: I'm trying that in the temporary patch [NOT MERGE] Test Jenkins, let's just see if we are lucky to get fullstack pass | 05:18 |
| janonymous | limao : Sure, thanks | 05:19 |
| *** diogogmt has quit IRC | 05:19 | |
| *** tonanhngo has joined #openstack-kuryr | 05:26 | |
| *** tonanhngo has quit IRC | 05:27 | |
| openstackgerrit | janonymous proposed openstack/kuryr-libnetwork: Add option to fetch kuryr uri from env https://review.openstack.org/363414 | 05:28 |
| limao | hi janonymous | 05:37 |
| limao | 2016-09-06 05:32:53.441914 | 2016-09-06 05:32:53.441 | Run: 4 in 34.090959 sec. | 05:37 |
| limao | 2016-09-06 05:32:53.444924 | 2016-09-06 05:32:53.444 | - Passed: 3 | 05:37 |
| limao | 2016-09-06 05:32:53.447828 | 2016-09-06 05:32:53.447 | - Skipped: 0 | 05:37 |
| limao | 2016-09-06 05:32:53.450597 | 2016-09-06 05:32:53.450 | - Failed: 1 | 05:37 |
| limao | 3 passed , 1 failed | 05:37 |
| limao | 2016-09-06 05:32:53.421022 | 2016-09-06 05:32:53.420 | docker.errors.APIError: 500 Server Error: Internal Server Error ("NetworkDriver.Join: vif_type(ovs) is not supported. A binding script for this type can't be found.") | 05:37 |
| limao | The failed one seems like because did not find bind script | 05:37 |
| janonymous | libexexec ones | 05:39 |
| *** diogogmt has joined #openstack-kuryr | 05:41 | |
| openstackgerrit | janonymous proposed openstack/kuryr-libnetwork: Docker Daemon to use port 2375 instead of 23750 at some places https://review.openstack.org/363414 | 05:42 |
| *** janki has joined #openstack-kuryr | 05:45 | |
| *** tonanhngo has joined #openstack-kuryr | 05:47 | |
| *** tonanhngo has quit IRC | 05:48 | |
| *** reedip has quit IRC | 05:50 | |
| openstackgerrit | janonymous proposed openstack/kuryr-libnetwork: Add option to fetch kuryr uri from env https://review.openstack.org/356357 | 05:55 |
| janonymous | limao: any leads ? | 05:56 |
| limao | hi janonymous, it should be because we use abs path for exec script in config file : /usr/libexec/kuryr/ , but we did not install the scripts in proper folder, it should be fixed in rpm package and devstack install scripts | 05:58 |
| *** reedip has joined #openstack-kuryr | 06:03 | |
| janonymous | Please find these small patches for review : https://review.openstack.org/#/q/owner:jaivish.kothari%2540nectechnologies.in+status:open++project:openstack/kuryr-libnetwork | 06:09 |
| *** tonanhngo has joined #openstack-kuryr | 06:41 | |
| *** tonanhngo has quit IRC | 06:43 | |
| apuimedo | good work guys! | 06:49 |
| apuimedo | I'll try to review them all today | 06:50 |
| openstackgerrit | janonymous proposed openstack/kuryr-libnetwork: Update home page link in cfg file https://review.openstack.org/365889 | 06:55 |
| openstackgerrit | Liping Mao proposed openstack/kuryr-libnetwork: Fix Rally test job in kuryr-libnetwork https://review.openstack.org/365851 | 06:56 |
| openstackgerrit | Liping Mao proposed openstack/kuryr-libnetwork: [NOT MERGE] Test Jenkins https://review.openstack.org/365683 | 06:57 |
| *** devvesa has joined #openstack-kuryr | 07:03 | |
| *** tonanhngo has joined #openstack-kuryr | 07:19 | |
| *** tonanhngo has quit IRC | 07:21 | |
| *** tonanhngo has joined #openstack-kuryr | 07:38 | |
| *** tonanhngo has quit IRC | 07:40 | |
| *** garyloug has joined #openstack-kuryr | 07:57 | |
| *** pablochacin has joined #openstack-kuryr | 07:57 | |
| *** tonanhngo has joined #openstack-kuryr | 08:08 | |
| *** tonanhngo has quit IRC | 08:09 | |
| *** salv-orlando has joined #openstack-kuryr | 08:29 | |
| *** salv-orl_ has quit IRC | 08:32 | |
| *** tonanhngo has joined #openstack-kuryr | 08:42 | |
| *** tonanhngo has quit IRC | 08:44 | |
| *** tonanhngo has joined #openstack-kuryr | 09:08 | |
| *** tonanhngo has quit IRC | 09:09 | |
| *** limao has quit IRC | 09:19 | |
| *** sdake has joined #openstack-kuryr | 09:45 | |
| *** reedip has quit IRC | 09:51 | |
| *** reedip has joined #openstack-kuryr | 10:04 | |
| *** tonanhngo has joined #openstack-kuryr | 10:23 | |
| janonymous | irenab: ping | 10:23 |
| *** tonanhngo has quit IRC | 10:23 | |
| irenab | janonymous: hi | 10:24 |
| janonymous | irenab: i think for these small commits there might not be need to file a bug/bp as these are minor changes or necessary things for kuryr to run... what do you say ..? | 10:25 |
| irenab | janfor the small commit that just improvment, I agree. but seems that its not just beautify, but fixing bugs or inconsistency. I prefer to have a way to trace them | 10:26 |
| janonymous | yes i agree, but like https://review.openstack.org/#/c/362642 and https://review.openstack.org/#/c/356357/ are modifications in existing code base... | 10:27 |
| irenab | what was the trigger to do it? | 10:27 |
| irenab | was it something that didn’t work? | 10:28 |
| irenab | it is just helpful when it has topic based on bug/bp | 10:28 |
| irenab | and just a suggestion, its prefered to use branch to work on the patch, not directly master ( I sometimes forget it though) | 10:29 |
| janonymous | for patch 362642 : Every project uses a single line to enable plugin , for patch 356357 : env vars for only URL were not supported in options, | 10:30 |
| irenab | janonymous: in general, we should follow http://docs.openstack.org/infra/manual/developers.html | 10:31 |
| janonymous | Ohh, you mean topic ? | 10:31 |
| irenab | janonymous: yes | 10:31 |
| irenab | topic branch | 10:32 |
| janonymous | irenab: thanks, i will change the patches as per requirements.. | 10:33 |
| *** sdake has quit IRC | 10:33 | |
| irenab | janonymous: sorry, for being so pedantic. Its for the good cause :-) | 10:33 |
| janonymous | irenab: No problem, every project have its own guidelines :) | 10:34 |
| *** ivc_ has joined #openstack-kuryr | 10:35 | |
| irenab | janonymous: indeed :-) | 10:36 |
| *** HenryG_ is now known as HenryG | 10:46 | |
| *** tonanhngo has joined #openstack-kuryr | 10:58 | |
| *** tonanhngo has quit IRC | 10:59 | |
| *** tonanhngo has joined #openstack-kuryr | 11:14 | |
| *** tonanhngo has quit IRC | 11:16 | |
| *** tonanhngo has joined #openstack-kuryr | 11:42 | |
| *** tonanhngo has quit IRC | 11:43 | |
| *** salv-orl_ has joined #openstack-kuryr | 11:52 | |
| *** salv-orlando has quit IRC | 11:52 | |
| *** reedip has quit IRC | 12:02 | |
| *** garyloug has quit IRC | 12:03 | |
| *** garyloug has joined #openstack-kuryr | 12:03 | |
| *** tonanhngo has joined #openstack-kuryr | 12:09 | |
| *** garyloug has quit IRC | 12:09 | |
| *** tonanhngo has quit IRC | 12:10 | |
| *** reedip has joined #openstack-kuryr | 12:16 | |
| *** oanson has quit IRC | 12:22 | |
| *** tonanhngo has joined #openstack-kuryr | 12:40 | |
| *** tonanhngo has quit IRC | 12:41 | |
| *** vikasc has quit IRC | 12:47 | |
| *** tonanhngo has joined #openstack-kuryr | 12:56 | |
| *** tonanhngo has quit IRC | 12:57 | |
| *** garyloug has joined #openstack-kuryr | 12:59 | |
| *** diga has quit IRC | 12:59 | |
| *** vikasc has joined #openstack-kuryr | 13:00 | |
| *** tonanhngo has joined #openstack-kuryr | 13:17 | |
| *** tonanhngo has quit IRC | 13:18 | |
| *** irenab has quit IRC | 13:31 | |
| *** irenab has joined #openstack-kuryr | 13:32 | |
| *** vikasc has quit IRC | 13:43 | |
| *** yuval has joined #openstack-kuryr | 13:55 | |
| *** vikasc has joined #openstack-kuryr | 13:55 | |
| *** yuval has left #openstack-kuryr | 13:55 | |
| *** huikang has joined #openstack-kuryr | 13:55 | |
| *** devvesa has quit IRC | 14:00 | |
| *** banix has joined #openstack-kuryr | 14:11 | |
| *** devvesa has joined #openstack-kuryr | 14:17 | |
| *** diogogmt has quit IRC | 14:20 | |
| *** janki has quit IRC | 14:26 | |
| *** salv-orlando has joined #openstack-kuryr | 14:29 | |
| *** salv-orl_ has quit IRC | 14:32 | |
| *** tonanhngo has joined #openstack-kuryr | 14:33 | |
| *** reedip is now known as reedip_outofoffi | 14:34 | |
| *** reedip_outofoffi is now known as reedip | 14:34 | |
| *** tonanhngo has quit IRC | 14:35 | |
| *** sdake has joined #openstack-kuryr | 14:44 | |
| *** devvesa has quit IRC | 15:10 | |
| *** diogogmt has joined #openstack-kuryr | 15:11 | |
| *** janki has joined #openstack-kuryr | 15:13 | |
| *** tonanhngo has joined #openstack-kuryr | 15:16 | |
| *** huikang has quit IRC | 15:16 | |
| *** huikang has joined #openstack-kuryr | 15:17 | |
| *** huikang has quit IRC | 15:22 | |
| *** devvesa has joined #openstack-kuryr | 15:25 | |
| *** sdake has quit IRC | 15:44 | |
| *** devvesa has left #openstack-kuryr | 15:56 | |
| *** huikang has joined #openstack-kuryr | 16:08 | |
| *** huikang has quit IRC | 16:12 | |
| openstackgerrit | Merged openstack/kuryr-libnetwork: Fix wrong dir in post_test_hook.sh https://review.openstack.org/365827 | 16:42 |
| openstackgerrit | Merged openstack/kuryr-libnetwork: Fix Rally test job in kuryr-libnetwork https://review.openstack.org/365851 | 16:42 |
| *** sdake has joined #openstack-kuryr | 16:57 | |
| *** huikang has joined #openstack-kuryr | 16:58 | |
| *** huikang has quit IRC | 17:04 | |
| *** huikang has joined #openstack-kuryr | 17:05 | |
| openstackgerrit | OpenStack Proposal Bot proposed openstack/kuryr-libnetwork: Updated from global requirements https://review.openstack.org/351976 | 17:06 |
| *** janki has quit IRC | 17:08 | |
| *** tonanhngo has quit IRC | 17:10 | |
| apuimedo | ivc_: did you have time to review the spec I sent? | 17:11 |
| ivc_ | apuimedo: yup | 17:12 |
| apuimedo | good! | 17:12 |
| apuimedo | any comments? | 17:12 |
| *** sdake_ has joined #openstack-kuryr | 17:12 | |
| apuimedo | :P | 17:12 |
| ivc_ | well i've got similar but a bit different idea about watchers/translators | 17:13 |
| ivc_ | i want to fit neutron events there and make it not just k8s specific but more generic | 17:14 |
| *** garyloug has quit IRC | 17:14 | |
| apuimedo | ivc_: could you please add the comment to gerrit, so it stays noted there | 17:15 |
| ivc_ | right now i'm in the process of deciding how i want to approach the implementation of it. i.e. some sort of observable or pub/sub or actor-ish model | 17:15 |
| *** sdake has quit IRC | 17:15 | |
| apuimedo | ivc_: what are you leaning towards? | 17:16 |
| ivc_ | as soon as i have to say something meaningful i'll certainly comment there. right now its just a bunch of shapeless ideas | 17:16 |
| apuimedo | cool | 17:18 |
| apuimedo | let me know if you want help | 17:18 |
| ivc_ | sure | 17:18 |
| apuimedo | ;-) | 17:18 |
| *** tonanhngo has joined #openstack-kuryr | 17:20 | |
| ivc_ | general the idea would be to have somewhat atomic 'actions' that would interact with neutron/k8s and be triggered by 'state events'. and the 'actions' would provide feedback by posting 'state event' which will then be picked by other actions | 17:21 |
| *** tonanhngo has quit IRC | 17:21 | |
| apuimedo | yup, that's similar to how I imagine what you mentioned | 17:22 |
| ivc_ | prolly that sounds gibberish | 17:22 |
| apuimedo | I imagined it more in an actor model btw | 17:22 |
| ivc_ | yeah i'd like to have some lifecycle control too | 17:22 |
| ivc_ | so we can stop/start/restart it | 17:22 |
| apuimedo | we definitely need that :P | 17:22 |
| apuimedo | I had this slightly crazy idea to publish neutron stuff to kubernetes/etcd | 17:23 |
| apuimedo | and watch those | 17:23 |
| ivc_ | that would add another syncing issue | 17:24 |
| ivc_ | i was thinking about a k8s-like watch api provider that would use neutron notifications that irenab suggested | 17:24 |
| apuimedo | well, I wanted to just push neutron notifications to k8s/etcd | 17:25 |
| *** huikang has quit IRC | 17:25 | |
| ivc_ | or more specifically mimicking k8s watch concept (i.e. that you get the current state as 'added' events when you subscribe) | 17:25 |
| *** huikang has joined #openstack-kuryr | 17:25 | |
| apuimedo | ivc_: heh | 17:25 |
| apuimedo | I thought that a bit ambitious | 17:26 |
| apuimedo | but it would be nice to have | 17:26 |
| apuimedo | ivc_: wouldn't that have to be a mini server that | 17:28 |
| apuimedo | a) received the neutron notifications | 17:28 |
| apuimedo | b) took as snapshot on start | 17:28 |
| ivc_ | yup. thats just an option, but for now we can stick with simple polling | 17:28 |
| *** sdake_ has quit IRC | 17:28 | |
| apuimedo | c) kept an in-memory model that it modified with the notifications | 17:29 |
| *** sdake has joined #openstack-kuryr | 17:29 | |
| ivc_ | not exactly like that | 17:29 |
| apuimedo | ivc_: that is probably doable | 17:29 |
| apuimedo | how? | 17:29 |
| ivc_ | dont need to keep in memory | 17:29 |
| ivc_ | if noone subscribed - just discard the notification | 17:29 |
| *** huikang has quit IRC | 17:30 | |
| ivc_ | when someone subscribes - start monitoring the neutron notifications and at the same time query api for the current state | 17:30 |
| *** huikang has joined #openstack-kuryr | 17:30 | |
| *** huikang has quit IRC | 17:30 | |
| ivc_ | i think thats how k8s folks do it | 17:30 |
| *** huikang has joined #openstack-kuryr | 17:31 | |
| apuimedo | interesting | 17:31 |
| apuimedo | and when a watcher ends, you end as well, or you keep processing with a timeout? | 17:31 |
| ivc_ | its pretty simple | 17:31 |
| ivc_ | yes. it would just end | 17:31 |
| apuimedo | cause it is normal for watchers to reconnect | 17:31 |
| apuimedo | well, it can always be optimized | 17:32 |
| apuimedo | simpler to just end indeed | 17:32 |
| apuimedo | I take note | 17:32 |
| * apuimedo going to trello | 17:32 | |
| ivc_ | yup. tho we'll have to go through the api query again. so if we reconnect every 5 seconds we'll kill neutron api server | 17:32 |
| apuimedo | right | 17:33 |
| apuimedo | but your point stands | 17:33 |
| apuimedo | optimizations can always come later | 17:33 |
| ivc_ | also got a suggestion for ha | 17:34 |
| apuimedo | ivc_: I wonder if that is not something we should eventually contribute to Neutron | 17:34 |
| *** huikang_ has joined #openstack-kuryr | 17:34 | |
| apuimedo | go ahead, please | 17:34 |
| ivc_ | so we use something like tooz+etcd driver to manage group membership between the raven nodes | 17:35 |
| *** huikang has quit IRC | 17:35 | |
| apuimedo | I'm not familiar with tooz | 17:36 |
| apuimedo | but I'll check it out | 17:36 |
| ivc_ | does not matter really. just some way to check the state of the cluster and monitor join/leave | 17:36 |
| apuimedo | 17 | 17:36 |
| ivc_ | 17? | 17:36 |
| apuimedo | sorry, weechat not registering a command | 17:36 |
| apuimedo | :-) | 17:36 |
| apuimedo | ivc_: the plan was using etcd for the cluster membership, yes | 17:37 |
| ivc_ | but anyway. the point is that each node has its ordering number which we use to filter events | 17:37 |
| ivc_ | so if the node has n=3 and we have k=10 nodes we filter events where uid%k == n | 17:38 |
| ivc_ | and buffer all other events | 17:38 |
| ivc_ | on each cluster heartbeat we check the cluster state. if nothing changed we discard the buffer | 17:38 |
| ivc_ | if nodes joined/leaved we reevaluate uid%k == n with new k/n for buffered messages and proccess them when needed | 17:39 |
| ivc_ | pretty much like that | 17:40 |
| apuimedo | ivc_: we have to care about the events that were not attended due to ravens dying | 17:40 |
| ivc_ | not really | 17:40 |
| ivc_ | when we restart the watcher we get all the events anyway | 17:41 |
| apuimedo | that's right | 17:41 |
| ivc_ | i mean we can set ?watch=true&timeout=180 and then every 3 minutes we'll have the current state anyway | 17:42 |
| apuimedo | but let's say you are downsizing the RC of raven | 17:42 |
| apuimedo | ivc_: that's cheating :P | 17:42 |
| ivc_ | that works :P | 17:42 |
| apuimedo | it means having to process the whole state every X minutes | 17:42 |
| ivc_ | we can trigger it if we see the cluster state change | 17:42 |
| ivc_ | does not need to happen on some regular interval | 17:43 |
| apuimedo | right | 17:44 |
| apuimedo | if we see the cluster downsize we can make any member start anew | 17:44 |
| apuimedo | so it covers missing ground | 17:44 |
| * apuimedo taking notes | 17:45 | |
| ivc_ | i like it because its dumb as a brick and keeps things simple enough | 17:45 |
| apuimedo | those are two of my favorite design principles | 17:46 |
| ivc_ | tho i'm a little worried about the late events. but thats something we'll have to deal anyway with active-active | 17:46 |
| apuimedo | ivc_: you mean the latency to process events of a dead member? | 17:48 |
| ivc_ | i mean events comming out of order | 17:48 |
| ivc_ | and not reflecting the real state of the world | 17:49 |
| apuimedo | like endpoints after services? | 17:49 |
| *** banix has quit IRC | 17:49 | |
| *** sdake has quit IRC | 17:49 | |
| ivc_ | like port being up posted to the buffer, getting deleted and then the message get processed | 17:49 |
| ivc_ | so we are notified that port is up but the port is not there by that moment | 17:50 |
| ivc_ | not really specific to ha. could happen even with single instance | 17:50 |
| apuimedo | ivc_: the filters can help a bit with that | 17:51 |
| *** banix has joined #openstack-kuryr | 17:51 | |
| ivc_ | not exactly filters as i see it but kind of yes | 17:51 |
| ivc_ | to me filters are cpu-bound and sync | 17:52 |
| *** huikang_ has quit IRC | 17:53 | |
| apuimedo | I find it legitimate for a filter to check if something exists on Neutron, like for example a port not existing for a pool member | 17:54 |
| *** huikang has joined #openstack-kuryr | 17:54 | |
| apuimedo | and do something like discard the event in that case | 17:54 |
| ivc_ | i just have a different pipeline for that kind of events | 17:55 |
| ivc_ | but the idea is the same | 17:55 |
| *** sdake has joined #openstack-kuryr | 17:56 | |
| apuimedo | my idea for out of order events was mostly to discard them | 17:56 |
| apuimedo | and then, for late events that needed them | 17:56 |
| apuimedo | generate constructed events for the potentially missing stuff | 17:57 |
| apuimedo | let's say you get first the endpoint events and later teh service event | 17:57 |
| apuimedo | you discard the endpoint events | 17:57 |
| ivc_ | yes. for k8s we can keep the current resourceVersion and discard anything below it | 17:57 |
| apuimedo | and then, when you process the service event | 17:57 |
| ivc_ | but for neutron its not as simple | 17:57 |
| ivc_ | > generate constructed events | 17:58 |
| apuimedo | you do a GET for its endpoints and insert the events, or in HA, you do some bullshit change to trigger events | 17:58 |
| ivc_ | i'd rather avoind taking that path | 17:58 |
| apuimedo | (from k8s) | 17:58 |
| apuimedo | you could add metadata to have k8s send the events | 17:58 |
| apuimedo | it's kind of "constructed" | 17:58 |
| apuimedo | :-) | 17:58 |
| ivc_ | its hacky | 17:58 |
| apuimedo | I won't deny that | 17:59 |
| apuimedo | how would you treat this case? | 17:59 |
| apuimedo | delaying the endpoint events? | 17:59 |
| apuimedo | then you risk event bufferbloat | 17:59 |
| apuimedo | but it is certainly an option if well managed | 18:00 |
| ivc_ | oh that kind of events | 18:00 |
| *** huikang has quit IRC | 18:00 | |
| *** banix has quit IRC | 18:00 | |
| *** huikang has joined #openstack-kuryr | 18:01 | |
| ivc_ | i've thought about that case yesterday | 18:02 |
| ivc_ | somehow i've decided that its not an issue and will work find but right now can't remember why. | 18:02 |
| ivc_ | *fine | 18:03 |
| apuimedo | well, even with a single raven we experienced this (when raven didn't have locks) | 18:03 |
| *** banix has joined #openstack-kuryr | 18:03 | |
| *** huikang_ has joined #openstack-kuryr | 18:03 | |
| *** huikang has quit IRC | 18:05 | |
| apuimedo | (and locks are not a solution I'd like to have after the upstreaming) | 18:06 |
| ivc_ | agreed | 18:06 |
| ivc_ | well i'll get to that anyway when i'll get to that part of code | 18:08 |
| *** huikang_ has quit IRC | 18:09 | |
| *** huikang has joined #openstack-kuryr | 18:09 | |
| apuimedo | ivc_: :-) | 18:09 |
| apuimedo | let me know when you get there | 18:09 |
| apuimedo | I'm now checking about the functional testing options | 18:10 |
| *** oanson has joined #openstack-kuryr | 18:10 | |
| ivc_ | btw | 18:13 |
| apuimedo | tell me | 18:13 |
| ivc_ | https://github.com/ivc/kuryr-poc-demo | 18:13 |
| ivc_ | yeah was getting a link :) | 18:13 |
| ivc_ | could salvage hyperkube launcher from there | 18:14 |
| *** huikang has quit IRC | 18:14 | |
| *** huikang has joined #openstack-kuryr | 18:15 | |
| ivc_ | more specifically https://github.com/ivc/kuryr-poc-demo/blob/master/roles/k8s/templates/kubelet.default.j2 | 18:15 |
| ivc_ | and manifests from https://github.com/ivc/kuryr-poc-demo/tree/master/roles/k8s/templates | 18:16 |
| apuimedo | oh, that's what devvesa meant! | 18:16 |
| apuimedo | I don't generally use manifests :P | 18:16 |
| ivc_ | its for api-server/etcd | 18:16 |
| apuimedo | I just use hyperkube to start api-server, etcd and others | 18:17 |
| ivc_ | i like it how you only need to care about the kubelet and let it start apiserver | 18:17 |
| apuimedo | but with docker run commands | 18:17 |
| apuimedo | ivc_: I prefer being explicit in general :P | 18:17 |
| ivc_ | yeah but then it uses its own manifests | 18:17 |
| ivc_ | oh you mean you start hyperkube api manually right | 18:18 |
| ivc_ | anyway the command is in the manifest | 18:19 |
| ivc_ | does not really matter if we use manifests to run it or run it ourselves | 18:19 |
| apuimedo | sure, the important thing is that it is an option | 18:20 |
| apuimedo | with minikube... I thought about either using the generic SSH driver | 18:20 |
| apuimedo | (since it uses docker machine, it could be possible, though it requires careful checking) | 18:21 |
| apuimedo | or the openstack driver | 18:21 |
| apuimedo | but hyperkube seems easier at this point | 18:21 |
| apuimedo | ivc_: btw, the documentation for running them manually https://github.com/midonet/kuryr/blob/k8s/doc/install-guide/source/cloud-config-k8s-controller.yaml | 18:22 |
| ivc_ | yeah i've seen that :P | 18:22 |
| apuimedo | I like that then you don't get the kubernetes containers listed inside kubernetes, so testing code can be simpler | 18:22 |
| apuimedo | ivc_: I was trying to find it on contrib/demo, but they moved it :P | 18:22 |
| apuimedo | sometimes I think that people in kubernetes have allergy to writing function docs | 18:24 |
| ivc_ | i'd avoid using minikube. hyperkube is simple enough and fast | 18:24 |
| apuimedo | ivc_: agreed 100%. But since I agreed to check its feasibility in the weekly IRC meeting, I was doing the checks ;-) | 18:24 |
| apuimedo | so then we can give a reasoning for the decision | 18:24 |
| *** oanson has quit IRC | 18:29 | |
| *** tonanhngo has joined #openstack-kuryr | 18:32 | |
| *** tonanhngo_ has joined #openstack-kuryr | 18:34 | |
| *** tonanhngo_ has quit IRC | 18:36 | |
| *** tonanhngo has quit IRC | 18:36 | |
| *** tonanhngo has joined #openstack-kuryr | 18:37 | |
| *** huikang has quit IRC | 19:01 | |
| *** huikang has joined #openstack-kuryr | 19:02 | |
| *** banix has quit IRC | 19:02 | |
| *** huikang_ has joined #openstack-kuryr | 19:05 | |
| *** pablochacin has quit IRC | 19:05 | |
| *** huikang has quit IRC | 19:06 | |
| *** banix has joined #openstack-kuryr | 19:20 | |
| *** huikang_ has quit IRC | 19:28 | |
| *** huikang has joined #openstack-kuryr | 19:29 | |
| *** diogogmt has quit IRC | 19:29 | |
| *** diogogmt has joined #openstack-kuryr | 19:31 | |
| *** vikasc has quit IRC | 19:42 | |
| *** huikang has quit IRC | 19:43 | |
| *** huikang has joined #openstack-kuryr | 19:44 | |
| *** vikasc has joined #openstack-kuryr | 19:44 | |
| *** sdake has quit IRC | 19:46 | |
| *** huikang_ has joined #openstack-kuryr | 19:47 | |
| *** huikang has quit IRC | 19:48 | |
| apuimedo | I just sent an email to the ML about the functional testing for kuryr-kubernetes | 19:54 |
| apuimedo | after we discuss we can move to bp and impl | 19:54 |
| *** sdake has joined #openstack-kuryr | 20:03 | |
| *** huikang_ has quit IRC | 20:03 | |
| *** huikang has joined #openstack-kuryr | 20:04 | |
| *** huikang has quit IRC | 20:08 | |
| *** huikang has joined #openstack-kuryr | 20:10 | |
| *** banix has quit IRC | 20:11 | |
| *** salv-orl_ has joined #openstack-kuryr | 20:30 | |
| *** salv-orlando has quit IRC | 20:33 | |
| *** pablochacin has joined #openstack-kuryr | 20:44 | |
| *** sdake has quit IRC | 21:02 | |
| *** huikang has quit IRC | 21:02 | |
| *** diogogmt has quit IRC | 21:45 | |
| *** banix has joined #openstack-kuryr | 21:48 | |
| *** banix has quit IRC | 21:49 | |
| *** banix has joined #openstack-kuryr | 21:50 | |
| *** banix has quit IRC | 22:00 | |
| *** banix has joined #openstack-kuryr | 22:01 | |
| *** banix has quit IRC | 22:39 | |
| *** ivc_ has quit IRC | 23:07 | |
| *** sdake has joined #openstack-kuryr | 23:11 | |
| *** pablochacin has quit IRC | 23:17 | |
| *** hongbin has quit IRC | 23:17 | |
| *** diga has joined #openstack-kuryr | 23:41 | |
| *** tonanhngo has quit IRC | 23:55 | |
Generated by irclog2html.py 2.14.0 by Marius Gedminas - find it at mg.pov.lt!