15:56:37 <Sai> #startmeeting MD-SAL Interest Call 15:56:37 <odl_meetbot> Meeting started Tue Mar 29 15:56:37 2016 UTC. The chair is Sai. Information about MeetBot at http://ci.openstack.org/meetbot.html. 15:56:37 <odl_meetbot> Useful Commands: #action #agreed #help #info #idea #link #topic #startvote. 15:56:37 <odl_meetbot> The meeting name has been set to 'md_sal_interest_call' 15:56:44 <Sai> #chair colindixon 15:56:44 <odl_meetbot> Current chairs: Sai colindixon 15:57:05 <colindixon> hey 15:57:08 <Sai> #topic Previous Meeting Minutes 15:57:42 <Sai> #link - https://meetings.opendaylight.org/opendaylight-meeting/2016/md_sal_interest_call/opendaylight-meeting-md_sal_interest_call.2016-03-22-16.02.html 15:57:49 <Sai> #Topic Agenda 15:58:52 <Sai> #info - 1. Call for any topics to be added to the agenda that people didn't get around to adding. 15:59:00 <Sai> #info 2. Review the existing action items 15:59:24 <Sai> #info 3. Bug 5504 - Controller crashes with OOM. 15:59:37 <Sai> #info 4. Karaf Upgrades 15:59:48 <Sai> #info 5. Blueprint Update 16:05:40 <Sai> #topic Review the existing action items 16:06:12 <Sai> #info - Ryan & Bala will link the Bug (behavioral constraints in Be 5484) to Bug 5396 - Done 16:06:24 <Sai> #info - Tony will sync up with Peter wrt Bug 5396 - Done 16:06:34 <Sai> #info - Dynamic Filter Injection Documentation - Done 16:06:45 <Sai> #link - https://wiki.opendaylight.org/view/AAA:DynamicFilterFramework 16:07:07 <Sai> #topic - Bug5504 Controller crashes with OOM. 16:07:17 <Sai> #link - https://bugs.opendaylight.org/show_bug.cgi?id=5504 16:07:28 <Sai> #info - itr was a blocker at one point for beryllium SR1 16:08:17 <Sai> #link - Ryan asks if it is fixed in master ? 16:08:22 <Sai> #info - Ryan asks if it is fixed in master ? 16:08:41 <Sai> #info - it is done 16:09:32 <Sai> #info - Robert says old/root is no more there. 16:09:41 <Sai> #info - changed to toString() 16:10:19 <Sai> #info - TomP has an idea the next steps wrt BluePrint 16:11:50 <Sai> #info - Followup patch exists in git 16:12:16 <Sai> #info - git patch Change id 36812 16:12:39 <Sai> #link - https://git.opendaylight.org/gerrit/#/c/36812/1 16:13:11 <Sai> #info - TomP sees it happening if config yang file fails to push 16:15:05 <Sai> #Action - TomP opens a bug to track the potential bug ( config yang change - fails to push - reproducible in local box ) 16:15:17 <Sai> #topic - Karaf Upgrades 16:15:52 <Sai> #info - Stephen says he prefers Karaf 4.0, but it breaks a bunch of code 16:16:33 <Sai> #info - Stephen says he needs to figure out how to deal with karaf 3.0 16:16:54 <Sai> #info - Ryan has a patch for KARAF-4457 in upstream karaf 3.X stream to update pax-web 16:17:06 <Sai> #link - https://issues.apache.org/jira/browse/KARAF-4457 16:17:30 <Sai> #info - Ryan is Working with Jamie Goodyear and team to get some movement here 16:17:49 <Sai> #info - probably not until .6 or .7 release 16:19:08 <Sai> #info - TomP refers to startup properties , requests Ryan to check with stephen 16:19:25 <Sai> #info - Stephen says we can do it without waiting for karaf release 16:20:30 <Sai> #info - TomP says empty one install all features, we are missing admin bundle (which BluePrint needs). ODL karaf installs everything but no the the empty one. 16:20:54 <Sai> #info - TomP asks is there any preference for ODL karaf over empty one ? 16:21:57 <Sai> #info - TomP says changing karaf configurations is not a really good solution 16:23:18 <Sai> #info - We need not upgrade entire containers, since karaf team is good at having revision based upgrades 16:24:13 <Sai> #Action - Ryan to followup with TomP regarding upstream karaf 3.X stream to update pax-web 16:24:59 <Sai> #Info - Stephen says rebuilding the distribution is current issue with karaf 4.X 16:25:29 <Sai> #info - Stephen says it is time consuming since we need to figure out each build fail 16:25:59 <Sai> #info - Ryan volunteers to help Stephen 16:27:41 <Sai> #topic - Blueprint Update 16:29:59 <Sai> #info - TomP poiont to change id 35867 and wades through the blueprint implementation details 16:30:44 <Sai> #info - TomP explains the toaster example in context with BluePrint 16:35:17 <Sai> #info - TomP emphasis on proxy mechanism and why it is double edged sword 16:36:41 <Sai> #info - TomP - points to "restart dependant on updates" - it walks the service hierarchies - same as config system like atomic restarts 16:38:33 <Sai> #info - TomP says there is a default for all reference type, (default means user will land up in CDS) 16:42:33 <Sai> #info - TomP actually implemented config system & Blueprint side to side and got them working !!!! 16:43:08 <Sai> #TomP says toaster is also compatible with config system 16:43:55 <Sai> #undo 16:43:55 <odl_meetbot> Removing item from minutes: <MeetBot.ircmeeting.items.Info object at 0x24cfa10> 16:44:00 <Sai> #info - TomP says toaster is also compatible with config system 16:47:15 <Sai> #info - AutoColsable in Remote RPC broker Module might be a good example for users to check out the different between both 16:47:41 <Sai> #info - TomP says it works with both Restconf & Netconf 16:50:23 <Sai> #info - Blueprint theoretically should be possible to do with same feature config in the order of the feature hierarchy 16:53:16 <Sai> #info - Implicit dependencies in blueprint are done via service or blueprint advertiser. it makes possible to add an extension which solves the dependencies problem. (think about OVSDB dependent on OF) 16:54:22 <Sai> #Action - TomP will make sure to have a wiki on how Blueprint is implemented / how it works 16:55:31 <Sai> #info - TomP waiting on patch "Improve config system shutdown" to be reviewed by Robert / Tony 16:58:36 <Sai> #info - TomP says he is waiting on code reviews from other committers in the controller project 16:59:51 <Sai> #info -Robert says he would be very careful wrt to config admin (single point of failure) when blueprint is completely switches over to. 17:01:16 <Sai> #info - TomP says it is little harder to do right now, but have plans to implement 17:02:47 <Sai> #info - TomP wades through toaster-provider-impl in change id 36405 - showing the user facing config & actual code wiring. 17:05:08 <Sai> #info - TomP says in short we have to make user config system as an option to do user operation 17:06:35 <Sai> #info - colin says we have to more consolidate the user interfering way and how much we should allow user to make changes. 17:06:50 <colindixon> Sai: thanks for taking the notes 17:07:38 <Sai> #info - Robert says Netty configs using the datastore implementation should be considered 17:08:02 <Sai> #info - Robert says using sharding we can overcome some issues 17:08:26 <colindixon> #info rovarga says what worries him the most is the fragmentation of configuration state 17:08:32 <Sai> #info Robert says this is excellent progress but only he is worried about framentation of configuration state 17:09:37 <Sai> #info - Robert says devices don't have yang representation , and when we ask everyone to jump on to blueprint .. there may be a problem since we don't have ready made configs 17:10:11 <colindixon> #info is there a compelling reason to store configuration outside the data store 17:10:26 <Sai> #info - TomP says use config system side by side , bridge netconf to config admin are the possible options 17:10:39 <colindixon> #action rgoulding to add things about user config and making sure we have ideally one option to the next week's agenda 17:10:55 <Sai> #info thanks colindixon 17:11:08 <Sai> #endmeeting