#acumos-meeting: Archtitecture Committee

Meeting summary

1. Issues and blockers (farheen_cefalu_a, 14:11:22)
   1. ML Workbench is blocked. Integration of the nifi jupyter into AIO. Bryan is required on the call. (farheen_cefalu_a, 14:12:13)
   2. Kazi - there are 8 apis that need to be called to create the flow in registry and they work successfully in unsecured nifi. 7/8 apis are successful with security. Bryan join the call. (farheen_cefalu_a, 14:14:30)
   3. Bryan I would like the experts to step in but no one is on the call. We have PTLs and team members. Maybe they can jump in and help. Bryan - the harm is that my time is better spent in working on the issues. We can get some kubernetes experts if we escalate. (farheen_cefalu_a, 14:17:43)
   4. Bryan - Sharing screen Nifi registry having trouble with in store value. The nifi registry has to serve the user and the pipeline service in a secure manner to be used by user, pipeline, and admin services. The method to do the secure mode is to create a key store and trust store and provision the with those. I have experience with key and trust stores. This is a shell output of the component as its runni (farheen_cefalu_a, 14:20:40)
   5. something that you can help with. Running the script the key tool command was able to decipher and open the key store and verifies that its valid. Once the application comes up its an apache application that does a lot of initialization stuff and at some point it gives key store password incorrect it's a java jeti app server. If anyone has information why this would work. Please raise their hand. Please (farheen_cefalu_a, 14:25:18)
   6. Bryan send Mukesh the log files. Bryan - Once that registry problem is fixed and we have defined pipeline certificate then we can start testing the admin experience of managing the buckets then test programatically the users. Each of these lines has to be verified. Vaibhav did jupyter work for Docker adopting to kubernetes will require work. I don't know how the notebook service is using the apis. I have (farheen_cefalu_a, 14:28:35)
   7. tested. I would like help from Vaibhav or anyone else. Another critical path issue. I need the portal to deploy the api. Mukesh - done. It needs to be tested. We dropped deployment of pipeline with the k8 branch. (farheen_cefalu_a, 14:30:39)
   8. The entire team thought that pipeline was dropped. Deployment under kuberentes is dropped. Bryan sharing the Boreas acumos diagram. We dropped kafka, flink, and pipeline in a private kubernetes. We are not going to be deploying pipelines as running artifacts in the k8 environment. (farheen_cefalu_a, 14:32:46)
   9. Bryan is slogging security verification. Requires Mukesh's assistance. Making a small change in the jar needs the whole portal. This is not agile. We have to back away from embedding jars. It is a bad architecture decision Mukesh will try to put a debug statement in the package. Willt ry to do it today if possible. (farheen_cefalu_a, 14:38:21)
   10. Bryan we have a number of things that have to be done including understanding how these artifacts have to be retrieved which is different. We don't have time to figure out another solution. Kazi In the design time the user will be able to all the things. Design time nifi will be available. Will the end user be able to use the nifi pipeline at design time? (farheen_cefalu_a, 14:42:29)
   11. Bryan it will be useful like jupyter. We can write a user guide of how to use the pipeline manually. The missing pieces are from I want to deploy this to the pipeline component (farheen_cefalu_a, 14:45:07)
   12. This was envisioned back in January and knew. Once we moved away from kubeflow and moved towards nifi we depended on the CMLP team to assist. (farheen_cefalu_a, 14:48:50)
   13. Nat suggested working on a call with Steve Polston later this afternoon. Kazi - we want to spend our time testing. (farheen_cefalu_a, 14:54:03)
   14. endmeeting (farheen_cefalu_a, 14:54:12)

Action items

1. (none)

People present (lines said)

1. farheen_cefalu_a (17)
2. collabot (3)