Page History

2017-12-08 OW Tech Interchange - Meeting Notes

• New member on Dragos’ team from AdobeIO

• Main/core OpenWhisk
  • Matt: please review quarterly report draft
  • Carlos: highlight a few updates for ~ last 3 weeks
    • CLI source code removed from main OW repo. No manual syncs, all resources in CLI and client-go repo.”
    • REST api not used for testing instead of CLI
    • Jeremias may want to discuss kafka
    • CouchDB in cluster mode submitted/merged from at-large contribution
    • Simplified config. of Scala components (to JVM)
    • Some updates to LogStore, ability to stream to file
    • Trigger fires used to be async, something broke, now 202 return code (accepted)
    • Some additional contribs around Alarm feed (triggers), fire once plus start/stop date.  Can now get config info from it.
    • removed swift 3.0, now 3.1.1 is default Swift 4 being worked
    • NodeJS 8 now avail. (in addition to NodeJS 6)
    • Controller HA enabled by default (clustered)
    • Cleaned up old DB “views”
    • New deploy package repo. created from Belinda, uses wskdeploy, helps with DevOps
  • Dave Grove: Kube update
    • Dave shared presentation
    • Recent Kube depl. updates 
      • Re-synced all thing with upstream OW project
      • can now use minikube (default for travis now), docs being updated to reflect this
      • Reduced # of custom images
      • Now creating a Kube CouchDB image
      • Added ingress configuration + some examples
      • Merged Daemonset for invokers (from Daniel)
      • Enabled HA for controllers, stallin on Couch/kafka
    • Focus areas going forward:
      • Upstream switch to pureconfig, should reduce build breakage when config changes
        • 4 breaks last week for example
      • Complete deployment (package install, route mgmt)
      • Kafka+Zookepper HA
      • Run OW test suite on Kube
  • Rob: can you post about the #110 issue? Dave yes, will do
  • Carlos: If Red Hat using Kube depl. it is good to have another OW Kube provider, need 2nd ref. impl. of OW (not just IBM)
  • Priti: wskdeploy and project sync (managed)
    • Pres. shared
    • What is Whisk Deploy?  
      • To be part of whisk.system actions
      • manifest.yaml read and deploys all from
      • managed is a new feature, syncs client and server packages based upon a calculated hash
        • logic flow chart shared to show how comparison.sync done
      • demo of feature
  • Priti: second demo.  Enhancement requested to support dependencies in NodeJS with ZIP (at server)
    • currently, it is a manual process, NPM install and ZIp at client
    • this new action (server-side) can do this for you.
    • Supports NodeJS and Python now (via packages.json and requirements.txt respectively)
    • This new Package is called “build” in whisk.system
    • demo given of new Action using NodeJS and Python "helloworld" example and “jokes” from James' blog
  • Chetan - Apache Maven
    • Trying to work on SPI for storage layer
    • 2 parts required, having maven artifacts, start w. the snapshots (so they can be referenced for compilation)
    • 
      
      • and configuration
    • Submitted 2 PRs for these things
    • Could use Apache snapshot repo., then have proper releases of these as part of overall release process
    • After PR merged, can try push of binaries to repo.
    • Carlos: Markus made same comment, should be for tests too.  Other repos. could benefit from this.  Can we share this code?
    • Chetan: PR will include tests
    • Carlos: prob. do this from Jenkins, not Travis as we are exceeding 45 minute build/test mark.  If we can move to Jenkins that is better. The VMs in Jenkins already have access to write to Apache.
    • Chetan: sent an email around on how queries are handled on CouchDB, made some suggestions to support Mongo (more declarative than computed) and other DBs.  Please review.

• Matt volunteers. for Jan 3rd meeting (skipping Dec. 20th meeting)