You are viewing a plain text version of this content. The canonical link for it is here.
Posted to commits@airflow.apache.org by "Ash Berlin-Taylor (JIRA)" <ji...@apache.org> on 2017/11/27 21:37:00 UTC
[jira] [Issue Comment Deleted] (AIRFLOW-1852) Allow hostname to be
overridable
[ https://issues.apache.org/jira/browse/AIRFLOW-1852?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]
Ash Berlin-Taylor updated AIRFLOW-1852:
---------------------------------------
Comment: was deleted
(was: Where is the hostname currently used in airflow? (I've been running fine without worrying about this, as I'm sure are lots of other people.)
{quote}
Since the web server calls out to the individual worker nodes to snag logs, what happens if one dies midway?
{quote}
There's support for writing task logs to GCS or S3 for more persistent storage.)
> Allow hostname to be overridable
> --------------------------------
>
> Key: AIRFLOW-1852
> URL: https://issues.apache.org/jira/browse/AIRFLOW-1852
> Project: Apache Airflow
> Issue Type: Improvement
> Reporter: Trevor Joynson
>
> * https://github.com/apache/incubator-airflow/pull/2472
> This makes running Airflow tremendously easier in common
> production deployments that need a little more than just
> a bare `socket.getfqdn()` hostname for service discovery
> per running instance.
> Personally, I just place the Kubernetes Pod FQDN (or even IP) here.
> Question: Since the web server calls out to the individual
> worker nodes to snag logs, what happens if one dies midway?
> I may later look into that, because that scares me slightly.
> I feel like workers should not ever hold such state, but that's purely a personal bias.
> Thanks,
> Trevor
--
This message was sent by Atlassian JIRA
(v6.4.14#64029)