As always, they're available at https://docs.google.com/document/d/1JGmJrgeg98bHw_0_sSRmyX6WiAe13OdErcFlaz6Aa04/edit# For convenience, I have reproduced today's notes here:
Attendees: Zachary, Adam, Sarjeet, Brandon, Darin, JohnO, Miguel Intro: Zachary is an undergrad at Case-Western, already contributing code News/Updates: - MesosCon CFP is Closing today! - ApacheCon - Mohit and Adam to present Myriad in Vancouver in May - Hadoop Summit - looks like a submitted myriad talk is being voted on. - Strata? JohnO speaking about a different topic; suggest meetup on dev list [AI: John] - Brandon speaking about Mesos+Myriad at the end of the month, will share with dev@ - March Podling report: http://wiki.apache.org/incubator/March2016 - ‘mslackbot’ on IRC will alert Mesosphere employees if you just say their names (adam, mohit, kensipe) Agenda/Notes: - JohnO suggested an easier path (vagrant/docker images) for setting up a dev environment and playing around with it. Don’t want to have to install Mesos and Hadoop just to get Myriad running. - Bug fixes for earlier 0.1.1 release? - Multitenancy (0.2.0?) - Biggest issue: get FGS to work “well” - MYRIAD-153 <https://issues.apache.org/jira/browse/MYRIAD-153>: Placeholder tasks yarn_container_* is not cleaned after yarn job is complete. - MYRIAD-188 <https://issues.apache.org/jira/browse/MYRIAD-188>: Zero sized node managers can cause the Resource Manager to crash with an NPE - JohnO suggests a blog post promoting multitenancy and how it works; John is happy to do blogging/demos, but would use MapR. - Dockerized RM + NM (0.2.0?) - Darin: Don’t like dependencies (e.g. Java) on nodes, package them with Myriad. - Cautious excitement for the Mesos unified containerizer feature - Oversubscription (0.3.0?) - Could enable idle coarse-grained NM to offer its resources to other frameworks. - If preempting Myriad resources, would prefer a graceful shutdown that lets Myriad choose which containers/tasks to shut down. - IP per container (0.3.0?) - Calico module works with mesos containerizer already - 0.2.0 Release <https://issues.apache.org/jira/browse/MYRIAD-153?jql=fixVersion%20%3D%20%22Myriad%200.2.0%22%20AND%20project%20%3D%20MYRIAD%20AND%20resolution%20%3D%20Unresolved%20ORDER%20BY%20priority%20DESC> (end of Apr - mid May) - Miguel notices TeraSort jobs fail (and NMs die) on Myriad+MapR with 1 trillion rows; asks what testing we do. Darin suspects the default JVM overhead is too small. AM was on different node, so why won’t job continue with relaunched tasks? Darin will attempt to repro. Suggests we start a thread on dev@. Pull Requests <https://github.com/apache/incubator-myriad/pulls> and JIRAs <https://issues.apache.org/jira/issues/?jql=project%20%3D%20MYRIAD%20AND%20status%20in%20%28Open%2C%20%22In%20Progress%22%2C%20Reopened%29%20ORDER%20BY%20updated%20DESC> - MYRIAD-186 Clean up the build <https://github.com/apache/incubator-myriad/pull/60> - Zachary cleaned up after Darin’s comments; Darin to re-review - MYRIAD-180 Build should not pollute sources <https://github.com/apache/incubator-myriad/pull/59> - Zachary needs to investigate CI build error - Addresses Myriad 188 - NodeManager switch to UNHEALTHY causes NPE on … <https://github.com/apache/incubator-myriad/pull/62> - Zachary looked, but we need a full review. - This PR Addresses MYRIAD-153: tasks not finishing when FGS is enabled… <https://github.com/apache/incubator-myriad/pull/57> - Santosh took a look; need to merge
