[
https://issues.apache.org/jira/browse/FLINK-27341?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17629614#comment-17629614
]
Weihua Hu commented on FLINK-27341:
-----------------------------------
Hi, [~chesnay] , [~gaoyunhaii] I think this was introduced by FLINK-24474. It
uses the loopback address as the default address. In these configs, it only
supports flink clusters running on a single host, and taskmanager must use the
loopback interface to connect with the jobmanager, since the jobmanager only
binds the loopback interface. But if we don't set the bind-address to
localhost, taskmanager should not use the loopback interface to find its
external address. otherwise, this will cause other TaskManagers to not connect
with it.
IMO, we can determine whether to use the loopback interface by whether the
taskmanager.bind-host is loopback address.
If this change is acceptable, you can assign this to me.
> TaskManager running together with JobManager are bind to 127.0.0.1
> ------------------------------------------------------------------
>
> Key: FLINK-27341
> URL: https://issues.apache.org/jira/browse/FLINK-27341
> Project: Flink
> Issue Type: Bug
> Components: Runtime / Coordination
> Affects Versions: 1.15.0, 1.16.0
> Reporter: Yun Gao
> Priority: Major
>
> If some TaskManagers running with JobManager on the same machine while some
> other TaskManager not, the TaskManagers running together with JobManager
> would bind to localhost or 127.0.01, which makes the Netty connections across
> the TaskManagers fail.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
