[
https://issues.apache.org/jira/browse/SLIDER-91?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Steve Loughran resolved SLIDER-91.
----------------------------------
Resolution: Won't Fix
I propose fixing this in YARN-913 by making the service registry entry
persistent; (cleanable by YARN), with an ephemeral {{/live}} znode to indicate
live, and ephemeral component entries.
the service entry will outlive the AM, but the /live tag will indicate whether
or not the service is actually running at this point.
> slider registry code to gracefully handle (transient) ZK outages
> ----------------------------------------------------------------
>
> Key: SLIDER-91
> URL: https://issues.apache.org/jira/browse/SLIDER-91
> Project: Slider
> Issue Type: Sub-task
> Components: registry
> Affects Versions: Slider 0.30
> Reporter: Steve Loughran
>
> Slider registry code needs to handle transient registry failures of running
> clusters.
> we may want to fail hard on startup, but during normal operation, things
> shouldn't fail the AM
> # HTTP registry queries
> # any updates to registry status (log? queue?)
--
This message was sent by Atlassian JIRA
(v6.2#6252)
