flink-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Stephan Ewen <se...@apache.org>
Subject Re: a problem about flink cluster
Date Fri, 21 Aug 2015 15:48:50 GMT

Usually, TaskManagers should automatically choose a network interface from
which they can communicate to the JobManager. They also use that network
interface to talk to each other, then.

If I understand correctly, what happened is that the TaskManager that ran
on the same machine as the JobManager chose an interface that cannot
communicate with the other TaskManagers.

You can manually assign a TaskManager a hostname via the
"taskmanager.hostname" config value, but this needs to be in a config that
is exclusive to that TaskManager.


On Fri, Aug 21, 2015 at 5:08 PM, 华文 <404823056@qq.com> wrote:

> Hi,
> There is a problem occured during I was running the flink on cluster.
> I stared the cluster with 3 nodes(one of them as master node ran the
> jobmanager and taskmanager while others as worker nodes ran taskmanager).
> First, I didn`t notice that one of the worker nodes also started Docker
> and Docker used a virtual network card named "docker0" as the first network
> environment(I used command "ifconfig" to look up) with a IP address
> "A"(sorry for information sensitivity).
> And the IP "A" is not the real IP address of this worker node since the
> real IP address is "B".
> As a result, the taskmanager in this worker node registered to the
> jobmanager with the IP "A".
> So the flink throwed "connect refuse" when I ran a application on the
> cluster.
> Then I killed all process of Docker, did the "ifconfig docker0 down", put
> the real IP "B" to the first network and re-ran the cluster.
> But these didn`t work and the taskmanager still registered with the IP "A".
> So whether I can specify the IP "B" for the taskmanager to registered?
> Or is there any other way to fix this probleam?
> Greetings,
> Huang Wei

  • Unnamed multipart/alternative (inline, None, 0 bytes)
View raw message