[ 
https://issues.apache.org/jira/browse/AURORA-1556?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15092252#comment-15092252
 ] 

Joshua Cohen commented on AURORA-1556:
--------------------------------------

I think that could get confusing in the degenerate case where someone 
configures an http health checker but does not request *any* ports.

Also, we don't really have a concept of a "default" port, do we? Assuming I 
haven't glossed over that for years, I'm going to assume you mean the case 
where a task requests a single port not named "health" and configures a health 
checker, then the health checker should use whatever port was requested. I 
think that falls apart for the case where the task requests multiple ports, 
none of which are named "health" because it's non-deterministic which port we 
should use in that case?

Better to just be explicit about it?

If we *were* to default health checking to some requested port when no "health" 
port is requested, I think we'd need a deprecation cycle as this would be 
changing behavior that some have come to rely on (that is, not binding a 
"health" would disable health checks).

> Configuring an http health checker without binding a health port should be 
> considered invalid configuration
> -----------------------------------------------------------------------------------------------------------
>
>                 Key: AURORA-1556
>                 URL: https://issues.apache.org/jira/browse/AURORA-1556
>             Project: Aurora
>          Issue Type: Task
>          Components: Executor, Thermos
>            Reporter: Joshua Cohen
>            Priority: Minor
>
> Today if you configure an http health checker but don't bind a health port, 
> we do not perform any health checks. Arguably this is invalid configuration 
> and the task should be rejected. If you'd like to disable health checks (e.g. 
> for a devel task), then no health check config should be present.



--
This message was sent by Atlassian JIRA
(v6.3.4#6332)

Reply via email to