Attendees: Matt, Jeremias, Sandeep, Tyson, Ramki, Himvath, Martin, Carlos
Notes:
FYI, Many attending Serverlessconf NY so attendance will be light today.
Introductions of attendees
  • Ramki:  mgr. of an Adobe team in India, mainly working on runtime for OW.
  • Sandeep: from Bangalore on team with Ramki, just trying to get involved in community.
Open comments on status 
  • Main/core OpenWhisk
    • Big change, controller dist. data store
    • Jeremias: found use cases/needs of Mesos and Kube deployments.  Added new tests and was not able to find any regressions
    • Jeremias: would like to work with Tyson and others to make sure they (these alt. depl.) are happy with these changes
    • Tyson: we are coming into some issues withLoad Balancer where it does not respect the async nature of counters.  Could lead to unexpected sched. we sent something to email list, but getting more data and plan to continue discussion there.
    • Tyson: any other interesting PRs Jeremias
    • Jeremias; Rodric had a design doc. to split activations into multiple.  The PR is open, but not sure if its ready
    • Addresses hug number of activations we are seeing in production to improve how we reduce artifacts/storage under large loads.
    • Tyson: no other agenda items, nothing specific…
    • Tyson: some open PRs on Logstore extensions goal is to decouple log storage from log DB.  Jeremias indicates he ist testing some of the changes
    • Jeremias: yes, but its a bit more complex but there are ties to auth/login services… would affect existing deplyments.
    • Have been working on a prototype this week.  When Markus and others are back we will work on the SPI and post results and designs to dev list and Wiki once we understand what other changes may be needed.  Do not want to break users if possible.
    • Jeremias: once the API key is available and we can tag the docker container we may need to pass other parameters as well to support the CLI.  Again, will move the discussion to mailing lists
    • Tyson: initial Mesos integration PR is there now too.   A container-based factory, but without LogStore PR/API it may nnot be useful to review until that work is done.
    • Jeremias: will see activity from us on Logstore as it is absolutely necessary for us in production as well.
    • Jeremias: on MEsos, with Akka clusters, do we mean to ahve a deply. test/build running in open to test Akka clustering and prove it works with Mesos/Kube?
    • Tyson: talked to Markus on this… as with Splunk or other integrations… goal is to first add tests to verify API usage is correct and the API “contracts” are reliable.  Having tjhis tested on a “live” system is a much larger endeavor.  Not sure if it is poss. to have a large scale int. test using Travis or similar.
    • Tyosn: in near time would need test fom user/api level HTTP traffic is going out in an expected manner.  Does this explain how I bleive we intend t move forward
    • Jeremias: it is a larger question about testing Mesos/Kube to test SPIs and in general test if depl. steps work
    • Tyson: do not know how well we can automate this easily today with what we have
    • Jeremias: I will post question on mailing list and get more feedback from Dave Grove and Ben on how to move forward.
    • Tyson” CLI/API auth extensible is also a recent discussion.  Have a dedicated ext. auth provider.  A while back a proto. was submitted, but wanted everyone to know we intend to revisit that soon.
    • Martin: doing metrics using Log interfaces, have some basic stuff up and running, ooking at config side of things.  Was please we did not have to add too much code to get it working. Still a WIP
    • Tyson: would be great to have gen. way to run metrics. 
    • Martin: Also, easy to run it locally suing statsd
    • Tyson: Carlos joined
    • Carlos: have runtime updates… have been creating repos like crazy maybe 40+ n0w, thats a lot for open source project
      • 1 repo. created per runtime language plus generic docker SDK
      • Cobol etc. others can use this same approach and benefit
    • Carlos: migrated repo. with all commits/structure all “in tact” everything same, but all just in independent repos.  Created a basic README per as well.  Travis is there, Tests are there…
    • once these tests are passed we have pseudo-tags that we use and post the images to Docker hub… also trying to see how we annotate what went into each that image (towards release notes).
    • Carlos: 5 PRs, python, JS, Java etc. into the main repo. It should be transparent if it works… Main OW will be pulled form DockerHub now so Travis etc. builds of main OW repo. should be much faster
    • Carlos: still moving code out… Dragos was concerned with naming the repos/images… but for now we do not want to rename as this might break people… can revisit later, but do not want people to experience this change and pulling from Docker hub integrated in Mesos/Kube before we make more potentially breaking (name) changes.
    • Carlos: this should help people better support versions of language runtimes e.g., Java 9, Swift 4? Should help everyone better maintain runtimes
    • Carlos: for some reason, Serverlessconf… people tweeting that PHP support is well received… so any way we can make people understand we can support ANY runtime language is good.
    • Tyson: this will do well and agree it is cool
Solicit topics for next call from attendees
  • None today
Confirm moderator for next call (i.e., Wed. Oct 25th 2nd)
  • Jeremias volunteers. 
  • No labels