05:01:32 #startmeeting servicevm-device-manager 05:01:33 Meeting started Tue Oct 7 05:01:32 2014 UTC and is due to finish in 60 minutes. The chair is yamahata. Information about MeetBot at http://wiki.debian.org/MeetBot. 05:01:34 Useful Commands: #action #agreed #help #info #idea #link #topic #startvote. 05:01:36 The meeting name has been set to 'servicevm_device_manager' 05:01:38 hello 05:02:11 #topic Announcement 05:02:11 hi 05:02:29 neutron rc1 is released 05:02:58 seems neutron-spec is opened. I suppose it will be announced neutron irc meeting today. 05:03:23 I updated WIP reference routervm implementation 05:03:32 #link https://review.openstack.org/#/c/116773/ wip routervm plugin 05:04:01 I also uploaded a patch to simplify csr1kv 05:04:08 #link https://review.openstack.org/#/c/125580/ simplify csr1kv 05:04:38 yamahata: I'm still reviewing your patches. 05:04:39 I created the patch by only reading the code, so it need review by CSR1kv guys. 05:04:46 bmelande: thanks. 05:05:28 bmelande: My direction is to make routervm implementation closer to builtin router implementation like l3 agent. 05:05:53 In order to clear implementation wart 05:06:24 That's all from me. Any other announcement? 05:07:00 bmelande: I have a question on cfg agent. I'd like to discuss it later. 05:07:16 ok, yes we dan do that after this 05:08:18 s3wong: any update on summit session? 05:08:25 yamahata: could you add some more details to what you meant by makng the routervm closer to builtin? 05:08:37 yamahata: no 05:09:06 hareeshpc: Sure 05:09:11 #topic Open Discussion 05:09:11 yamahata: i meant any specific points 05:09:12 yamahata: also have not seen any update from tox on the ML, so probably not finalized yet 05:10:00 hareeshpc: I'm trying to generalize CSR1kv implementation so that it can be used for other routervm. 05:10:32 So far CSR1kv implementation has its own code and has many deviation from l3 agent (or l3 ha or dvr) 05:11:01 Such deviations causes routervm implementation ugly and so on. 05:12:04 If the routervm plugin can be near as much as possible to l3 agent, the implementation would be simple/clean. 05:12:29 yamahata: is not because of the fact that implementation aspects are different for both? 05:12:41 For that purpose, I'm trying to simplify/refactor l3_db and csr1kv code. 05:14:39 yamahata: i think the cfg-agent is more similiar to l3-agent except in the configuration parts 05:14:47 hareeshpc: In principle yes. I'd like to make the code clean before complicating it. 05:15:10 yamahata: completly agree 05:15:16 hareeshpc: that's the point. I think the code can be more simplified. 05:15:23 or consolidated. 05:15:36 yamahata: your-message 05:16:33 So far I uploaded some refactoring/cleanup patch. and I'm planning some more. 05:17:07 yamahata: ok, i will take a look. thanks 05:18:49 any other issues? 05:19:34 okay, I'd like to discuss on cfg agent. 05:19:48 bmelande: I'm reading cfg agent and router service code. 05:20:04 yamahata: ok 05:20:12 l3 agent has polling code for message loss/reorder. 05:20:19 Sepcifically _process_routers_loop method. 05:20:37 But I don't find counter part of cfg agent/router service. 05:20:57 Do I miss it? or it isn't addressed yet? 05:21:42 yamahata: the equivalent is process_services() 05:22:34 It syncs cfg agent with neutron db. 05:22:54 What if routervm itself is not in sync with cfg agent? who fix it? 05:23:44 all the service specific code has been moved to a service-helper, in this dase a routing-service-helper 05:24:34 the sync happens there, customized for each service. that is the idea 05:24:38 hareeshpc: I'm talking about RoutingServiceHelper.process_service() 05:25:31 hareeshpc: Okay, I got the idea. 05:26:11 thanks 05:27:25 any other topics to discuss? 05:27:30 yamahata: about routervm itself not in sync with cfg agent, are you talking about the VM instance out of synch with agent? 05:28:04 bmelande: That's right. 05:28:37 cfg agent issues rest call to the VM. The rest call can be lost. 05:29:08 yamahata: i will check. but the service helper collects update messages for a period (cfg-agent op interval) and sync with neutron db for those 05:29:59 yamahata: ok. i understood now 05:30:55 yamahata: the rest call if it fail should throw an exception 05:31:06 In l3 agent case, the agent knows the actual state of routing table etc.. , so it doesn't matter. 05:32:37 okay, anything else to discuss? or time is becoming up. 05:33:36 yamahata: we can take this offline, so i know i git your point 05:33:48 git -> got 05:33:54 hareeshpc: sure. 05:34:08 thanks everyone. see you next week. 05:34:35 #endmeeting