| openstackgerrit | Merged openstack/dragonflow master: Remove useless parameters for super class of RyuDFAdapter https://review.openstack.org/458377 | 00:58 |
|---|---|---|
| *** dmellado has quit IRC | 01:00 | |
| *** dmellado has joined #openstack-dragonflow | 01:02 | |
| openstackgerrit | hujie proposed openstack/dragonflow master: remove duplicated bum flow for provider app https://review.openstack.org/459053 | 02:31 |
| *** VenuReddy1 has joined #openstack-dragonflow | 03:41 | |
| *** VenuReddy has quit IRC | 03:43 | |
| openstackgerrit | Merged openstack/dragonflow master: Fix mech_driver not found https://review.openstack.org/456137 | 03:49 |
| *** VenuReddy has joined #openstack-dragonflow | 04:18 | |
| *** VenuReddy2 has joined #openstack-dragonflow | 04:20 | |
| *** VenuReddy1 has quit IRC | 04:20 | |
| *** VenuReddy1 has joined #openstack-dragonflow | 04:21 | |
| *** VenuReddy has quit IRC | 04:24 | |
| *** VenuReddy2 has quit IRC | 04:24 | |
| *** VenuReddy has joined #openstack-dragonflow | 04:31 | |
| *** VenuReddy1 has quit IRC | 04:33 | |
| *** VenuReddy1 has joined #openstack-dragonflow | 04:43 | |
| *** VenuReddy has quit IRC | 04:46 | |
| *** gongysh has joined #openstack-dragonflow | 04:46 | |
| *** gongysh has quit IRC | 04:49 | |
| *** VenuReddy1 has quit IRC | 05:06 | |
| *** VenuReddy has joined #openstack-dragonflow | 05:06 | |
| *** VenuReddy1 has joined #openstack-dragonflow | 05:57 | |
| *** VenuReddy has quit IRC | 05:59 | |
| *** VenuReddy1 has quit IRC | 06:04 | |
| *** VenuReddy has joined #openstack-dragonflow | 06:04 | |
| *** VenuReddy has quit IRC | 06:17 | |
| *** VenuReddy has joined #openstack-dragonflow | 06:17 | |
| *** VenuReddy1 has joined #openstack-dragonflow | 06:28 | |
| *** VenuReddy has quit IRC | 06:28 | |
| *** l_jacob_ has quit IRC | 06:35 | |
| openstackgerrit | jingting proposed openstack/dragonflow master: Support VM live migration https://review.openstack.org/394759 | 06:40 |
| openstackgerrit | jingting proposed openstack/dragonflow master: Support VM live migration https://review.openstack.org/394759 | 06:42 |
| *** dimak_ has joined #openstack-dragonflow | 07:00 | |
| *** lewo has joined #openstack-dragonflow | 07:16 | |
| *** VenuReddy has joined #openstack-dragonflow | 07:17 | |
| *** VenuReddy1 has quit IRC | 07:19 | |
| *** VenuReddy1 has joined #openstack-dragonflow | 07:21 | |
| *** VenuReddy2 has joined #openstack-dragonflow | 07:23 | |
| *** VenuReddy has quit IRC | 07:23 | |
| *** VenuReddy1 has quit IRC | 07:25 | |
| oanson | dimak_, ping | 07:27 |
| dimak_ | oanson, yes | 07:27 |
| oanson | I | 07:27 |
| oanson | I'm looking at https://review.openstack.org/#/c/459638/3/dragonflow/ovsdb/impl_idl.py@96 | 07:27 |
| *** VenuReddy2 has quit IRC | 07:27 | |
| *** VenuReddy has joined #openstack-dragonflow | 07:28 | |
| oanson | I'm trying to think of a better flow that db_change_callback | 07:28 |
| dimak_ | Yes, look at patch above | 07:28 |
| dimak_ | I remove that | 07:28 |
| oanson | Right | 07:28 |
| dimak_ | I can squash them if you'd like | 07:28 |
| oanson | 1 sec, Let me look | 07:28 |
| oanson | I see. | 07:29 |
| oanson | But now the events are emitted on the wrong thread | 07:29 |
| oanson | db_change_callback pushes the event onto a queue, which apply_db_changes then pulls and acts | 07:29 |
| oanson | The changed code is in the ovsdb-monitor thread. | 07:30 |
| *** VenuReddy1 has joined #openstack-dragonflow | 07:30 | |
| dimak_ | We'll thats not a NB event... | 07:30 |
| oanson | apply_db_changes runs in the controller threads | 07:30 |
| dimak_ | Well* | 07:30 |
| oanson | True. But we want it to run in the controller thread, not the ovsdb-monitor thread | 07:30 |
| dimak_ | Other thing that bothered be is dragonflow.db.models.ovs.OvsPort location, its not a Db model.. | 07:31 |
| dimak_ | We keep calling it that and treating it as such but maybe thats wrong | 07:31 |
| oanson | You can move it to dragonflow.ovsdb | 07:31 |
| oanson | (A module inside, not directly on dragonflow.ovsdb (: ) | 07:32 |
| *** VenuReddy has quit IRC | 07:32 | |
| dimak_ | I think I'll do that | 07:32 |
| oanson | I'm thinking maybe we can pass a callable to the queue, have apply_db_changes call the callable | 07:32 |
| oanson | With action = 'callable' or something, like 'log' | 07:33 |
| dimak_ | And probably rename it | 07:33 |
| dimak_ | no no no | 07:33 |
| oanson | It was only a thought | 07:33 |
| dimak_ | Hide table/action/key/value inside a callable as well then | 07:34 |
| dimak_ | We can keep NbApi as the part that creates callables that apply db changes, and make local controller a thread that pops callables and executes them | 07:36 |
| dimak_ | Just an idea | 07:36 |
| oanson | That's what happens now | 07:37 |
| oanson | Just in api_nb, not df_controller | 07:37 |
| oanson | I'm beginning to think that the way it is at the end of 459638 is the best we can do | 07:37 |
| oanson | We don't need special handling in api_nb - that's done in df_controller | 07:38 |
| dimak_ | I just dislike the fact that we use table_name for a model that has no table | 07:38 |
| dimak_ | and applying it as if were a nb update | 07:39 |
| dimak_ | Its a bit confusing | 07:39 |
| oanson | mf.get_model works also on model names, not just table names, right? | 07:39 |
| dimak_ | Yeah, I though of changing that | 07:39 |
| oanson | So we can pass the model name to the event - then the model doesn't need a table name | 07:40 |
| dimak_ | Maybe we can change ModelBase.emit to push all events to a queue | 07:40 |
| dimak_ | then you can emit from any thread | 07:40 |
| dimak_ | and have a dispatcher thread to take care of the queue items | 07:40 |
| oanson | I like that 'emit' emits immediately | 07:40 |
| oanson | Maybe add a 'queue event' somewhere, saying we want the event to be emitted but by someone else | 07:41 |
| dimak_ | We can pass class name as table name and it'll work | 07:41 |
| dimak_ | its just awkward that I have to pack this thing into a DbChange objects | 07:41 |
| dimak_ | Plus jsonize and unjsonize it | 07:41 |
| oanson | Yes - that is strange | 07:42 |
| oanson | I like the dispatcher idea | 07:43 |
| oanson | Have db_change_callback tell the dispatcher to queue the relevant event | 07:44 |
| oanson | Then have OVSDB monitor also queue different events | 07:44 |
| dimak_ | Yes | 07:44 |
| oanson | Move the whole thing out of nb_api. The subscriber should queue events anyway | 07:44 |
| dimak_ | It's what it already does :P | 07:45 |
| oanson | Yes, but it does it through code in nb_api | 07:45 |
| *** VenuReddy has joined #openstack-dragonflow | 07:45 | |
| *** VenuReddy2 has joined #openstack-dragonflow | 07:46 | |
| oanson | dimak_, to summarize - we need a dispatcher that queues events, and then emits them on the correct thread | 07:48 |
| *** VenuReddy1 has quit IRC | 07:48 | |
| dimak_ | Why not handle all the events on a single thread / pool? | 07:49 |
| *** VenuReddy1 has joined #openstack-dragonflow | 07:49 | |
| oanson | The events are sent to the dispatcher by subscriber and OVSDB monitor. They are taken from the dispatcher by the controller | 07:49 |
| oanson | dimak_, yes. That's what the df-controller does | 07:49 |
| dimak_ | Yeah | 07:49 |
| oanson | In the bgp service case (xiaohhui's patches) the BGP service (not the DF controller) takes the events | 07:49 |
| *** VenuReddy has quit IRC | 07:50 | |
| dimak_ | We can add a "dispatcher/worker" facility | 07:50 |
| oanson | Yes. | 07:50 |
| dimak_ | And have any service invoke it on its last line of run() | 07:50 |
| *** VenuReddy2 has quit IRC | 07:50 | |
| oanson | Let's skip the thread-pool, however. That won't work well in python due to the GIL | 07:51 |
| oanson | dimak_, or a mixin overriding run. The publisher service, e.g., might use a different mechanism | 07:51 |
| dimak_ | Yes, we should iron out the details, see what fits the existing use cases best | 07:52 |
| oanson | Sounds like a spec :) | 07:52 |
| dimak_ | I'll scribe something up | 07:53 |
| oanson | Cheers | 07:54 |
| dimak_ | Though do we block OvsPort migration for that? | 07:54 |
| oanson | No | 07:54 |
| oanson | But the skip nbapi for ovsport is blocked for that | 07:55 |
| dimak_ | OvsPort is not referenced by any other model so its nothing urgent | 07:55 |
| oanson | True. And it depends on lport migration, which is still giving us trouble | 07:56 |
| oanson | But I'd be happy to see this migration come to a happy end :) | 07:57 |
| oanson | Commented on both reviews | 07:59 |
| *** VenuReddy has joined #openstack-dragonflow | 08:00 | |
| dimak_ | Great, thanks | 08:00 |
| *** VenuReddy2 has joined #openstack-dragonflow | 08:00 | |
| *** VenuReddy has quit IRC | 08:04 | |
| *** VenuReddy1 has quit IRC | 08:04 | |
| *** VenuReddy has joined #openstack-dragonflow | 08:04 | |
| *** VenuReddy1 has joined #openstack-dragonflow | 08:05 | |
| *** VenuReddy2 has quit IRC | 08:07 | |
| *** VenuReddy has quit IRC | 08:08 | |
| *** VenuReddy1 has quit IRC | 08:10 | |
| openstackgerrit | jingting proposed openstack/dragonflow master: Support VM live migration https://review.openstack.org/394759 | 08:20 |
| *** xiaohhui has joined #openstack-dragonflow | 08:24 | |
| *** VenuReddy has joined #openstack-dragonflow | 11:20 | |
| oanson | lihi, I've approved the SG migration patch. Note that there are some comments on it | 11:34 |
| oanson | Are you still planning a refactor to that feature? | 11:34 |
| oanson | lihi, Uhm, ping :) | 11:34 |
| lihi | oanson, pong :) | 11:35 |
| oanson | Just checking if you're here :) | 11:35 |
| lihi | yeah, but it will take some time until I'll get to it | 11:36 |
| oanson | No rush. As long as the functional stuff is in, no one is blocked | 11:36 |
| openstackgerrit | Omer Anson proposed openstack/dragonflow master: Support comparison of ModelProxy with None https://review.openstack.org/459626 | 11:41 |
| openstackgerrit | Merged openstack/dragonflow master: Migrate Security Groups to new model structure https://review.openstack.org/445548 | 13:08 |
| *** dimak_ has quit IRC | 15:17 | |
| *** VenuReddy1 has joined #openstack-dragonflow | 15:20 | |
| *** VenuReddy2 has joined #openstack-dragonflow | 15:20 | |
| *** VenuReddy has quit IRC | 15:23 | |
| *** VenuReddy1 has quit IRC | 15:25 | |
| *** VenuReddy has joined #openstack-dragonflow | 15:33 | |
| *** VenuReddy2 has quit IRC | 15:37 | |
| *** wangfeng has joined #openstack-dragonflow | 15:47 | |
| *** VenuReddy1 has joined #openstack-dragonflow | 16:02 | |
| *** VenuReddy has quit IRC | 16:05 | |
| *** dimak_ has joined #openstack-dragonflow | 18:58 | |
| *** saggi- has joined #openstack-dragonflow | 20:12 | |
| *** saggi has quit IRC | 20:13 | |
| *** saggi- is now known as saggi | 20:13 | |
| *** dimak_ has quit IRC | 23:10 | |
Generated by irclog2html.py 2.14.0 by Marius Gedminas - find it at mg.pov.lt!