14:10:37 #startmeeting Architecture Committee 14:10:37 Meeting started Wed Sep 4 14:10:37 2019 UTC. The chair is farheen_att. Information about MeetBot at http://wiki.debian.org/MeetBot. 14:10:37 Useful Commands: #action #agreed #help #info #idea #link #topic. 14:10:37 The meeting name has been set to 'architecture_committee' 14:10:50 #topic Acumos-1188 14:12:23 #info Bryan - We found that we don't have the flexibility to use k8 cluster. Host map is the only persistence volume of service. 14:13:19 #info not scalable so you have to distribute the components across a cluster node. A side car is related to that cluster of services. 14:14:11 #info this is an effort to deploy components across a k8 cluster. I am adding labels with their own log volumes. 14:14:44 #info I have a patch open to system integration to see the first stage. 14:15:14 #info Regarding the CI pipeline. The design of deployment is there. Still working on SV. 14:16:29 #info Looking at how the platform works with Docker in all aspects and consider all of those things are offloaded to an external service like Jenkins. It is a security concern. We are establishing a pattern. Onboarding team should take a look. 14:16:43 #info What is the scope in Clio? 14:17:47 #info Bryan - I don't think anything in Clio will be developed in Clio. There is an interplay of Jenkins and Acumos components. APIs need to be created and roles distributed. It is a significant aspect change across the platform. 14:18:30 #action Guy and Phillippe how does the onboarding code need to be refactored to fit into the CI/CD pipeline. What can be done in short term and what is coming up in the releases. 14:19:12 #info Guy - it is do-able but we may not have time to do it in Clio. It will disrupt and break things. 14:20:04 #info The ms code is missing. Even though at a high level we know what can be done and what is feasible. If you can bring together a plan we can review. 14:20:16 #action Manoop add this topic to next week's agenda. 14:20:37 #info Guy I agree we will discuss a bit and bring it up at the next architecture meeting. 14:21:41 #topic Access Control 14:23:10 #info Bryan the need for the Docker proxy is less and less therefore it's role can be simplified. Proposal to manage images through workflow that wraps the platform. 14:23:18 #topic HA Availability 14:24:32 #info Bryan this is a big ask. I don't know that the platform will scale. Suggested to plan out experiments to test whats possible. We need to take a planned experiment to see impacts. 14:24:59 #action Manoop Bryan wants to have a design review for HA 14:26:11 #info we can perform load tests to baseline what our platform can support. This will give us a clue to which components are impacted by the load tests. Save them in he HELM charts. 14:27:05 #info we need two action items. Load test and analyze the deployment techniques. 14:27:14 #info Ask the component leads these questions. 14:27:58 #action Ken to perform the load test to baseline the platform performance 14:28:23 #action Parichay to analyze the helm chart of k8 deployment to address HA/scalability 14:28:47 #info Is your component stateless? What aspects of your app depend on a state. What aspects are stateless? 14:29:53 #topic Search functionality 14:31:24 #info Many defects are being opened. Records per page. When we search we search on the 11 - 20 records. Search does not search all records. To do the entire database search a lot of changes will be required. 14:33:56 #info Do we really want to achieve the database search. Acumos-2940. 14:36:28 #action Tausif, Nat suggested collecting the search defects and we can review on the community call. 14:37:40 #action Tausif review with the product committee and then with the architecture committee. 14:40:17 #info regarding Acumos-1188 need support from the Federation team. When we do an un-publish on the model that has been federated we need to know the rules around how to handle a federated file. 14:41:49 #info when a model is federated from A to B. B wants to unpublish the model. If a model has been published and federated to another instance then the federated instance should un-publish. 14:42:05 #info un-publish will delete the model everywhere? 14:42:13 #info yes 14:44:49 #info Manoop once a model is federated it will show up somewhere. We can not remove a model once it is federated to the internet. Show the model as depricated. 14:45:31 #info what rites does the publisher have? 14:45:53 #info what rites does the publisher has after publication? 14:46:50 #info How instance A can changed ownership via instance B? Can we restrict further access? 14:47:02 #info It depends on license and terms if any. 14:48:08 #info Does the licensing team address the federated license models? 14:48:56 #info That has not been discussed. If the company wants to remove their model and they don't want anyone to use it they will have to do it offline. 14:49:18 #info Publisher takes the model down from A and B through the federation API? 14:49:33 #info yes, a scenario has to put together for that. 14:53:21 #info We can trigger the request from Portal front end but need Federation team to take the action for portal to trigger so it can be un-published. 14:54:49 #info this is not for-seen in the federation model. The need to withdraw a model we have to decide how to notify the peer. 14:55:35 #action Tausif put together a clear design and bring this topic to the Community call to Reuben. 14:58:20 #action Justin add link to documentation to meeting minutes to the documentation api 14:59:08 #action Manoop add Priya's Spark review to next weeks agenda. 15:00:32 #link https://wiki.acumos.org/display/AR/Architecture+Reviews+Score+Card 15:00:57 #info reviewed score card link above 15:01:07 #endmeeting