[
https://issues.apache.org/jira/browse/AURORA-1245?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14481300#comment-14481300
]
Steve Niemitz commented on AURORA-1245:
---------------------------------------
I disagree here. I think the current logic is more predictable. If the host
DNS is published at all, it should be constantly published every time.
Debugging why certain clients can't reach the scheduler sometimes, because one
scheduler decided to publish an IP address (which one?/what if it's possibly
unreachable externally?) would not be fun.
> Consider not resolving the server hostname at startup
> -----------------------------------------------------
>
> Key: AURORA-1245
> URL: https://issues.apache.org/jira/browse/AURORA-1245
> Project: Aurora
> Issue Type: Story
> Components: Scheduler
> Reporter: Kevin Sweeney
> Priority: Minor
>
> We currently abort startup if we're unable to resolve the scheduler hostname
> with a reverse DNS lookup. This seems like an unnecessary configuration
> burden - we could just use the local IP address if the reverse-lookup fails.
> We also publish this DNS name in ZooKeeper, but we could get away with
> publishing the the server IP address instead (clients that are finding us via
> ZooKeeper should be relying on ZooKeeper watch events to learn about changes
> to the scheduler IP address, not re-resolution of our DNS name every TTL).
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
