flink-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Till Rohrmann <trohrm...@apache.org>
Subject Re: Could not build up connection to JobManager
Date Thu, 05 Mar 2015 08:33:00 GMT
How did you start the flink cluster? Using the start-local.sh, the
start-cluster.sh or starting the job manager and task managers individually
using taskmanager.sh/jobmanager.sh. Could you maybe post the
flink-conf.yaml file, you're using?

With your changes, everything works, right?

On Thu, Mar 5, 2015 at 8:55 AM, Dulaj Viduranga <vidura.me@icloud.com>
wrote:

> Hi Till,
> I’m sorry. It doesn’t seem to solve the problem. The taskmanager still
> tries a 10.0.0.0/8 IP.
>
> Best regards.
>
> > On Mar 5, 2015, at 1:00 PM, Till Rohrmann <till.rohrmann@gmail.com>
> wrote:
> >
> > Hi Dulaj,
> >
> > I looked through your commit and noticed that the JobClient might not be
> > listening on the right network interface. Your commit seems to fix it. I
> > just want to understand the problem properly and therefore I opened a
> > branch with a small change. Could you try out whether this change would
> > also fix your problem? You can find the code here [1]. Would be awesome
> if
> > you checked it out and let it run on your cluster setting. Thanks a lot
> > Dulaj!
> >
> > [1]
> >
> https://github.com/tillrohrmann/flink/tree/fixLocalFlinkMiniClusterJobClient
> >
> > On Thu, Mar 5, 2015 at 4:21 AM, Dulaj Viduranga <vidura.me@icloud.com>
> > wrote:
> >
> >> The every change in the commit b7da22a is not required but I thought
> they
> >> are appropriate.
> >>
> >>> On Mar 5, 2015, at 8:11 AM, Dulaj Viduranga <vidura.me@icloud.com>
> >> wrote:
> >>>
> >>> Hi,
> >>> I found many other places “localhost” is hard coded. I changed them
in
> a
> >> better way I think. I made a pull request. Please review. b7da22a <
> >>
> https://github.com/viduranga/flink/commit/b7da22a562d3da5a9be2657308c0f82e4e2f80cd
> >>>
> >>>
> >>>> On Mar 4, 2015, at 8:17 PM, Stephan Ewen <sewen@apache.org> wrote:
> >>>>
> >>>> If I recall correctly, we only hardcode "localhost" in the local mini
> >>>> cluster - do you think it is problematic there as well?
> >>>>
> >>>> Have you found any other places?
> >>>>
> >>>> On Mon, Mar 2, 2015 at 10:26 AM, Dulaj Viduranga <
> vidura.me@icloud.com>
> >>>> wrote:
> >>>>
> >>>>> In some places of the code, "localhost" is hard coded. When it is
> >> resolved
> >>>>> by the DNS, it is posible to be directed  to a different IP other
> than
> >>>>> 127.0.0.1 (like private range 10.0.0.0/8). I changed those places
to
> >>>>> 127.0.0.1 and it works like a charm.
> >>>>> But hard coding 127.0.0.1 is not a good option because when the
> >> jobmanager
> >>>>> ip is changed, this becomes an issue again. I'm thinking of setting
> >>>>> jobmanager ip from the config.yaml to these places.
> >>>>> If you have a better idea on doing this with your experience, please
> >> let
> >>>>> me know.
> >>>>>
> >>>>> Best.
> >>>>>
> >>>
> >>
> >>
>
>

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