16:00:11 #startmeeting MD-SAL Interest Call 16:00:11 Meeting started Tue Apr 5 16:00:11 2016 UTC. The chair is Sai. Information about MeetBot at http://ci.openstack.org/meetbot.html. 16:00:11 Useful Commands: #action #agreed #help #info #idea #link #topic #startvote. 16:00:11 The meeting name has been set to 'md_sal_interest_call' 16:00:19 #chair colindixon 16:00:19 Current chairs: Sai colindixon 16:00:56 #topic last-meeting-minutes 16:00:59 #link https://meetings.opendaylight.org/opendaylight-meeting/2016/md_sal_interest_call/opendaylight-meeting-md_sal_interest_call.2016-03-29-15.56.html 16:01:09 #topic Agenda 16:02:02 #info 1. Call for any topics to be added to the agenda that people didn't get around to adding. 16:02:10 #info 2. Review the existing action items 16:02:18 #info 3.Infrastructure Issues and Compilation Problems this Week 16:02:26 #info 4. DCL Deprecation [3] 16:02:37 #info 5. Documentation of necessary changes to support YANG 1.1 [4] 16:02:51 #info 6. Update on Blueprint 16:05:10 #topic Action items 16:05:39 #info - Ryan to followup with TomP regarding upstream karaf 3.X stream to update pax-web - Active conversation going on 16:06:04 #info - some progress but we can't cross it off 16:06:29 #info - TomP opens a bug to track the potential bug ( config yang change - fails to push - reproducible in local box ) -- Merged to beryllium 16:07:06 #info - TomP will make sure to have a wiki on how Blueprint is implemented / how it works - Work in progress 16:07:56 #topic - Infrastructure Issues and Compilation Problems this Week 16:08:09 #info - cert problem was solved to use correct cert 16:08:17 #link - https://lists.opendaylight.org/pipermail/dev/2016-April/001874.html 16:08:31 #info - javax.xml.bind.UnmarshalException (s) 16:08:39 #link - https://lists.opendaylight.org/pipermail/controller-dev/2016-April/011830.html 16:09:10 #info - Ryan is not alone who is facing issue with xmls, Stephen too. 16:10:05 #info - colindixon asks to mail to release (Andy) , open a ticket using helpdesk. 16:11:42 #info - Tony says it seems xml being downloaded from internet, issue may be dealing with the place where we store. (it expects xml but it downloads html) 16:12:20 #info - Tony says target/features of individual repo will have more explanation on errors 16:12:47 #info we think they're really infrastructure failures, then we should report them to helpdesk or on #opendaylight-releng here 16:13:49 #info - colindixon says migration of nexus / other stuff shouldn't be ideally affecting these builds 16:14:43 #info colindixon says jenkins pulls from local mirror of nexus, if public nexus has any issues that explains it. 16:15:31 #Action - Ryan sends out mail / open ticket using helpdesk wrt javax.xml.bind.UnmarshalException (s) 16:16:10 #topic DCL Deprecation 16:16:23 #link - https://git.opendaylight.org/gerrit/#/c/36970/ 16:16:42 #info - Ryan says we need to deprecate the ACL deprecation as well 16:17:25 #Action - Ryan to followup with DCL Deprecation [3] 16:18:57 #info - Api's are used, Tony tries to deprecate all the old Api's so that it is easier for new changes / new Api 16:19:57 #info - Tony says If they are deprecated in Boron , they could be deleted in Boron. But with previous release experience it takes 2-3 releases in general 16:20:22 #info - Tony suggests to deal with deprecation after carbon 16:20:47 #info - colindixon supports it too. 16:21:17 #info - TomP suggests Nitrogen and Tony agrees. 16:22:04 #topic - Documentation of necessary changes to support YANG 1.1 16:22:11 #link - https://git.opendaylight.org/gerrit/#/c/37081/ 16:22:27 #info - Ryan requests for time frame 16:23:11 #info - The patch i just a small list, other part yang1.1 affects MD-SAl, given that we are close to M2, most likely in carbon. 16:23:44 #info - Tony says it introduces lot features , so it will be carbon. 16:24:15 #info - Ryan volunteers for any help here 16:24:52 #topic - Update on Blueprint 16:26:12 #info - TomP says - there are no new updates, bu there is progress in work wrt to apache karaf with help of Stephen 16:28:11 #info - Improve config subsystem Change 36777 - is a precursor and it is ready to go. 16:28:37 #info - TomP asks Tony/other commiters to review the patch 16:29:56 #info - TomP has several other patches to be merged and requests for volunteers to review / merge 16:30:42 #info - Stephen is interested, Tony requests to add Milos. 16:34:35 #Action - TomP will start to work on a wiki page to state how to move from config system to BluePrint 16:35:32 #info - Ryan offers help from AAA side. 16:36:04 #info - TomP suspects Netconf will be more harder wrt user configuration. 16:37:33 #info - Ryan asks if there are any people concerns , requests to raise sooner than later. 16:38:17 #topic - new topics 16:39:21 #info - question on patch 37009 wrt data tree consumers / data tree producers 16:40:12 #info - colindixon says it would be very confusing to developers, say OFP could be both consumer/producer in case of writing a flow . 16:41:46 #info - colindixon asks if there is any reason we can't do it in the context. 16:42:20 #info - Robert says we are throwing away standalone one way transactions. 16:43:18 #info - Robert says it is a consistency win as well as performace win. 16:44:27 #info - colin summarizes that data tree producer is a transacation factory and it is little bit more than that because it is a transaction chain 16:44:59 #info - colindixon requests not to introduce new terminology which will confuse developers 16:46:04 #info - Robert says since these are new concepts (not similar to data broker) we need to come up with new names. 16:46:44 #info - Robert is open for contribution 16:48:05 #info - Robert says with this patch we are moving readwritetransaction entirely 16:49:32 #info - colindixon says there are somethings (like increment values) are no more possible with this approach. 16:50:01 #info - Robert says we still have exclusive ability to write to a subtree 16:50:56 #info - Robert says we can have an another producer to write to subtree 16:51:54 #info - Muthu asks will the cardinality of 1:1 will be be enforced on a subtree 16:52:14 #info - Robert says Yes, At any given time only one producer can write to subtree. 16:52:50 #info - TOny says using locking mechanism it is possible. 16:53:28 #info - colindixon summarizes if a data tree is claimed by a procuder , you will be unable to write using netconf/restconf 16:54:49 #info - Robert says it is true, but it doesn't mean we can't negotiate mechanism, but it is out of the scope of these APIs 16:56:38 #info - colindixon says there are currently applications which usess single data tree where two producers are writing at same time. 16:57:19 #info - colindixon says in an application logical part & updation of status is an example 16:57:55 #info - Robert says intrinsic race condition comes in to picture. 16:59:53 #info - Robert says the main reason to do is pushing intent in to config and pushing status in to operational 17:00:26 #info - colindixon refers to difference between networking and Software way of looking at it. 17:01:39 #info - Robert says the system is never designed or architect-ed for these purposes 17:05:21 #info - colindixon says when write transaction is done it will fail when if sub tree is gone away or moved. 17:05:48 #info - Robert requests for use case / applications so that we can move ahead 17:07:23 #info - Tony rephrases Robert and asks is there is any application in odl which does listening / writing to the same subtree at same time. 17:08:12 #Action - colindixon to work on checking out an application in odl where it listens / writes to the same subtree at same time. 17:08:22 #topic - participants 17:09:54 #info - Sai / Ryan / Ajay Lele / Alexis / Andrej / Aolin / Martin Ciglan / Muthu / Robert /Shuva Kar / Stephen / Tom P / Tomas Cere / Tony / Vaclav / Viera / Vratko 17:09:58 #endmeeting