| openstackgerrit | Hong Hui Xiao proposed openstack/dragonflow: Revert "Start network connections after neutron forks workers" https://review.openstack.org/393013 | 01:02 |
|---|---|---|
| *** jtonsing has quit IRC | 01:07 | |
| openstackgerrit | Hong Hui Xiao proposed openstack/dragonflow: Revert "Start network connections after neutron forks workers" https://review.openstack.org/393013 | 01:16 |
| openstackgerrit | Merged openstack/dragonflow: Add license property to setup.cfg https://review.openstack.org/392590 | 01:50 |
| *** gongysh has joined #openstack-dragonflow | 02:59 | |
| *** oanson has joined #openstack-dragonflow | 03:23 | |
| openstackgerrit | WangJian proposed openstack/dragonflow: Allow to change chassis info after df controller restart https://review.openstack.org/365077 | 03:25 |
| rajivk | Hi, What happens if local controller goes down, how ovs bridge will get updated? | 03:53 |
| rajivk | can some other local controller take responsibility of configuring? | 03:53 |
| oanson | rajivk, no. Each controller is responsible for its local ovs bridge | 04:36 |
| oanson | if it goes down, then the bridge won't be updated until the local controller is restarted | 04:36 |
| rajivk | what about flows and traffic from existing VMs? | 04:37 |
| oanson | The flows won't change. | 04:37 |
| oanson | So connectivity for existing VMs (that don't need controller / use output:controller) will still have connectivity | 04:37 |
| rajivk | hmm, can we do something like, put all the controller information in db(RMC, redis etc) and keeps on monitoring them, if some controller goes down | 04:38 |
| rajivk | then other controller takes on control | 04:38 |
| oanson | rajivk, that's possible. But I think maybe a watchdog that restarts the local controller might be more in-line with our design | 04:39 |
| rajivk | okay, but i think, if local controller has gone down then there must be something, which caused it to go down, so may be restart will fail again. | 04:40 |
| rajivk | We can retry to some number of times and then pass on the control to other controller | 04:40 |
| rajivk | what do you think? | 04:40 |
| oanson | Should work. But only as a second fallback. A local watchdog should be an earlier line of defense | 04:43 |
| oanson | But it is better than a watchdog alone | 04:43 |
| rajivk | yes, i mean the same, watchdog try to restart it a few times(configurable) after that pass on the control if watchdog fails to restart | 04:43 |
| oanson | And having controllers inter-communicate could also have other benefits | 04:44 |
| rajivk | hmm | 04:44 |
| oanson | Just note that current dragonflow applications assume local controller for one local ovs bridge | 04:45 |
| oanson | That will have to be taken into account in the design, and possibly changed | 04:45 |
| rajivk | You are right. | 04:46 |
| rajivk | Do you think, we should discuss it in weekly meeting about it? | 04:46 |
| oanson | Sure | 04:47 |
| rajivk | What do you think of tracing each services of dragonflow for example, local controller, df-metadata service etc. Like we do in neutron, nova, heat , which will help to know the status of the whole cluster from a single point. | 04:47 |
| oanson | Sounds like a good plan. | 04:48 |
| oanson | I have to go. I'd be happy to discuss this again later today, or in the weekly meeting | 04:48 |
| oanson | You could also upload a spec for review | 04:48 |
| rajivk | okay, just ping me once you are online | 04:48 |
| oanson | Sure. Thanks. | 04:49 |
| *** openstackgerrit has quit IRC | 05:48 | |
| *** openstackgerrit has joined #openstack-dragonflow | 05:49 | |
| *** oanson has quit IRC | 06:08 | |
| *** irenab has joined #openstack-dragonflow | 06:46 | |
| *** oanson has joined #openstack-dragonflow | 06:54 | |
| *** irenab has quit IRC | 07:02 | |
| *** irenab has joined #openstack-dragonflow | 07:04 | |
| *** yuval has joined #openstack-dragonflow | 07:13 | |
| *** yuval is now known as Guest17054 | 07:14 | |
| *** Guest17054 is now known as yuval | 07:14 | |
| *** rajivk has quit IRC | 07:22 | |
| *** rajivk has joined #openstack-dragonflow | 07:34 | |
| *** dimak has joined #openstack-dragonflow | 07:43 | |
| *** openstackgerrit has quit IRC | 07:48 | |
| *** openstackgerrit has joined #openstack-dragonflow | 07:49 | |
| *** oanson has quit IRC | 08:13 | |
| *** oanson has joined #openstack-dragonflow | 08:15 | |
| *** oanson has quit IRC | 08:19 | |
| *** oanson has joined #openstack-dragonflow | 08:22 | |
| *** oanson has quit IRC | 08:22 | |
| *** oanson has joined #openstack-dragonflow | 08:23 | |
| *** oanson has quit IRC | 08:25 | |
| *** irenab has quit IRC | 08:27 | |
| *** oanson has joined #openstack-dragonflow | 08:28 | |
| oanson | rajivk, ping | 08:37 |
| rajivk | oanson, hi | 08:38 |
| oanson | I would like to continue the discussion from earlier | 08:38 |
| rajivk | ok, i have created blueprints for them | 08:39 |
| oanson | rajivk, great. Thanks! | 08:39 |
| oanson | It's this one, right? https://blueprints.launchpad.net/dragonflow/+spec/services-status | 08:40 |
| rajivk | yes | 08:40 |
| rajivk | What do you think, about consul added as a database? | 08:41 |
| rajivk | Consul has inbuilt support for health monitoring but i am not sure about pub-sub used in dragonflow. | 08:42 |
| oanson | More database plugins are great | 08:42 |
| oanson | The design of pluggability was to support as many as possible, according to the need of each deployer | 08:42 |
| oanson | If it doesn't have pub/sub, an external pub/sub can be used, e.g. the zmq driver | 08:43 |
| rajivk | okay | 08:43 |
| oanson | If you want to use its builtin health monitor support, then that also needs to be pluggable, so that if this feature is missing in other databases, they can be covered by other drivers | 08:43 |
| rajivk | ok | 08:44 |
| oanson | Or if this feature is supported by other databases, there's probably database-specific code that has to be written (e.g. different code for redis, consul, etcd) | 08:44 |
| rajivk | yes | 08:45 |
| rajivk | I have added blueprint for other topic we discussed as well | 08:46 |
| rajivk | https://blueprints.launchpad.net/dragonflow/+spec/ha-local-controller | 08:46 |
| rajivk | I don't know the way other services work except local controller, if they have the same issue then we can think of making them HA as well. | 08:47 |
| oanson | I think the local controller is the most important one | 08:47 |
| oanson | I think l3 behaves like in the reference implementation | 08:48 |
| oanson | metadata listens to connections. We can try to see how this can be load balanced. But is of much lesser priority than local controller | 08:48 |
| oanson | We should also have a user/admin notification mechanism in case of such (and other) failures. | 08:49 |
| oanson | But that's a separate blueprint | 08:49 |
| rajivk | I am not sure but notification(mail etc) has to be done in different component of openstack. | 08:49 |
| rajivk | Most of the time, we use different monitoring and notification service's not make out component responsible for them | 08:50 |
| oanson | yes, it would be better to interface with existing solutions/framework/api. | 08:50 |
| oanson | I don't think I understand your last message | 08:51 |
| rajivk | I mean in out product, we make our own monitoring and notification services. If there is already existing openstack component for notification and monitoring then we can think of integration with it. | 08:52 |
| rajivk | But if we implement monitoring and notification then we are deviating from purpose of dragonflow. | 08:52 |
| oanson | yes. | 08:52 |
| oanson | Dragonflow should be SDN/networking only. And monitoring/notification should be done in existing openstack infrastructure | 08:53 |
| rajivk | yes, that's what i mean. | 08:53 |
| oanson | We will have to interface with ceilometer, aodh, etc. | 08:53 |
| oanson | Then we agree :) | 08:53 |
| rajivk | Perfect :) | 08:53 |
| rajivk | oanson, thanks for the discussion. If i face any issue, i will ping you. | 08:54 |
| oanson | No problem. Thank you. | 08:55 |
| oanson | I am adding this issue to the agenda for Monday's weekly meeting | 08:55 |
| oanson | I hope you'll be able to be there and join the discussion | 08:55 |
| rajivk | thanks a lot | 08:55 |
| rajivk | On Monday, 9.00 AM UTC | 08:56 |
| rajivk | I will join it. :) | 08:56 |
| oanson | Yes. | 08:56 |
| oanson | Great! Thanks! :) | 08:56 |
| *** irenab has joined #openstack-dragonflow | 09:17 | |
| *** oanson has quit IRC | 09:31 | |
| *** oanson has joined #openstack-dragonflow | 09:32 | |
| *** oanson has quit IRC | 09:34 | |
| *** oanson has joined #openstack-dragonflow | 09:35 | |
| *** oanson has quit IRC | 09:37 | |
| *** oanson has joined #openstack-dragonflow | 09:38 | |
| *** oanson has quit IRC | 09:40 | |
| *** oanson has joined #openstack-dragonflow | 09:41 | |
| *** oanson has quit IRC | 09:43 | |
| openstackgerrit | WangJian proposed openstack/dragonflow: Allow to change chassis info after df controller restart https://review.openstack.org/365077 | 09:44 |
| *** oanson has joined #openstack-dragonflow | 09:44 | |
| *** oanson has quit IRC | 09:46 | |
| openstackgerrit | WangJian proposed openstack/dragonflow: Allow to change chassis info after df controller restart https://review.openstack.org/365077 | 09:47 |
| *** oanson has joined #openstack-dragonflow | 09:47 | |
| *** oanson has quit IRC | 09:49 | |
| *** oanson has joined #openstack-dragonflow | 09:50 | |
| *** oanson has quit IRC | 09:52 | |
| *** oanson has joined #openstack-dragonflow | 09:54 | |
| *** oanson has quit IRC | 09:55 | |
| *** oanson has joined #openstack-dragonflow | 09:56 | |
| *** oanson has quit IRC | 09:58 | |
| *** oanson has joined #openstack-dragonflow | 09:59 | |
| *** oanson has quit IRC | 10:01 | |
| *** oanson has joined #openstack-dragonflow | 10:02 | |
| *** oanson has quit IRC | 10:04 | |
| *** oanson has joined #openstack-dragonflow | 10:07 | |
| *** oanson has quit IRC | 10:07 | |
| *** gongysh has quit IRC | 10:57 | |
| openstackgerrit | WangJian proposed openstack/dragonflow: Allow to change chassis info after df controller restart https://review.openstack.org/365077 | 11:29 |
| openstackgerrit | Hong Hui Xiao proposed openstack/dragonflow: Small refactor of methods in dhcp_app https://review.openstack.org/393251 | 11:35 |
| openstackgerrit | Hong Hui Xiao proposed openstack/dragonflow: Re-download dhcp flow when enable dhcp of subnet https://review.openstack.org/393252 | 11:35 |
| *** saggi has joined #openstack-dragonflow | 12:11 | |
| *** oanson has joined #openstack-dragonflow | 12:12 | |
| *** lihi has joined #openstack-dragonflow | 12:26 | |
| *** oanson has quit IRC | 13:45 | |
| *** oanson has joined #openstack-dragonflow | 13:46 | |
| openstackgerrit | Hong Hui Xiao proposed openstack/dragonflow: Fix unable to update subnet when subnet disables dhcp https://review.openstack.org/393328 | 13:54 |
| *** lihi has quit IRC | 13:57 | |
| *** oanson has quit IRC | 14:05 | |
| *** irenab has quit IRC | 14:14 | |
| *** jtonsing has joined #openstack-dragonflow | 14:36 | |
| *** jtonsing has quit IRC | 14:52 | |
| *** jtonsing_ has joined #openstack-dragonflow | 14:52 | |
| openstackgerrit | Dima Kuznetsov proposed openstack/dragonflow: Forward DNAT traffic from table 7 https://review.openstack.org/392086 | 14:53 |
| openstackgerrit | Dima Kuznetsov proposed openstack/dragonflow: Refactor local_networks in L2ML2App https://review.openstack.org/392651 | 14:53 |
| *** oanson has joined #openstack-dragonflow | 14:59 | |
| *** yuval has quit IRC | 15:31 | |
| *** saggi has quit IRC | 15:34 | |
| *** jtonsing_ has quit IRC | 15:57 | |
| *** jtonsing has joined #openstack-dragonflow | 15:58 | |
| openstackgerrit | Merged openstack/dragonflow: Fix typo in L2ML2App https://review.openstack.org/392603 | 16:11 |
| openstackgerrit | Merged openstack/dragonflow: Updated from global requirements https://review.openstack.org/392722 | 16:12 |
| openstackgerrit | Merged openstack/dragonflow: Remove setting Q_L3_ENABLED https://review.openstack.org/392488 | 16:18 |
| openstackgerrit | Merged openstack/dragonflow: Disable q-meta when enable df-metadata https://review.openstack.org/392663 | 16:18 |
| openstackgerrit | Merged openstack/dragonflow: Remove df-ext-services https://review.openstack.org/391810 | 16:21 |
| openstackgerrit | Dima Kuznetsov proposed openstack/dragonflow: Nitpicks: Remove redundant pass https://review.openstack.org/393392 | 16:22 |
| openstackgerrit | Dima Kuznetsov proposed openstack/dragonflow: Nitpicks: Fix typo in RedisPortStatusNotifier https://review.openstack.org/393393 | 16:22 |
| openstackgerrit | Dima Kuznetsov proposed openstack/dragonflow: Nitpicks: Fix invalid attribute access https://review.openstack.org/393394 | 16:22 |
| *** dimak has quit IRC | 16:33 | |
| *** oanson has quit IRC | 18:27 | |
| *** oanson has joined #openstack-dragonflow | 18:38 | |
| *** oanson has quit IRC | 20:17 | |
| *** jtonsing_ has joined #openstack-dragonflow | 20:44 | |
| *** jtonsing has quit IRC | 20:45 | |
| *** jtonsing_ is now known as jtonsing | 20:45 | |
Generated by irclog2html.py 2.14.0 by Marius Gedminas - find it at mg.pov.lt!