You are viewing a plain text version of this content. The canonical link for it is here.
Posted to user@predictionio.apache.org by Tamas Soltesz <ta...@codeandsoda.hu> on 2019/05/03 14:50:36 UTC
PredictionIO in docker - pio train fails
Hello guys,
I am not sure about the process of asking questions here, so please be
gentle with me :)
I've asked my question on SO before I found this mailing list, would you
please take a look there?
https://stackoverflow.com/questions/55967253/pio-train-fails-with-ioexception-connection-reset-by-peer
Issue in short: I am running pio in docker swarm, and pio train fails with
IOException: Connection reset by peer _after_ calculating and saving the
model. I don't know how to find out what went wrong as I can't see any
suspicious in any of the logs.
If posting SO links here violates any rule here, please let me know. I can
post my question here as well separately if needed.
Thank you all,
Tamas Soltesz
Re: PredictionIO in docker - pio train fails
Posted by takako shimamoto <ch...@gmail.com>.
Maybe the following setting is incorrect:
PIO_STORAGE_SOURCES_ELASTICSEARCH_HOSTS: predictionio_elasticsearch
The correct host name is probably `elasticsearch`.
Regards,
Takako
On Tue, May 7, 2019 at 4:54 PM Tamas Soltesz
<ta...@codeandsoda.hu> wrote:
>
> Hi,
>
> Thank you for the reply. When not providing the --master local[num] parameter, it submits the train to the spark cluster. (see the MASTER env. variable in the docker compose file.)
> Now it seems like the exception is because the connection to the elasticsearch is lost. Currently I am testing this idea (by not using the elasticsearch at all).
>
> BR,
> Tamas Soltesz
>
> On Tue, May 7, 2019 at 8:08 AM takako shimamoto <ch...@gmail.com> wrote:
>>
>> Hi Tamas,
>>
>> Have you thought about submitting a training task to Spark Cluster?
>>
>> pio-docker train -- --master spark://spark-master:7077
>>
>> See: https://github.com/apache/predictionio/tree/develop/docker#run-with-spark-cluster
>>
>> On Fri, May 3, 2019 at 11:50 PM Tamas Soltesz
>> <ta...@codeandsoda.hu> wrote:
>> >
>> > Hello guys,
>> >
>> > I am not sure about the process of asking questions here, so please be gentle with me :)
>> >
>> > I've asked my question on SO before I found this mailing list, would you please take a look there?
>> > https://stackoverflow.com/questions/55967253/pio-train-fails-with-ioexception-connection-reset-by-peer
>> >
>> > Issue in short: I am running pio in docker swarm, and pio train fails with IOException: Connection reset by peer _after_ calculating and saving the model. I don't know how to find out what went wrong as I can't see any suspicious in any of the logs.
>> >
>> > If posting SO links here violates any rule here, please let me know. I can post my question here as well separately if needed.
>> >
>> > Thank you all,
>> > Tamas Soltesz
Re: PredictionIO in docker - pio train fails
Posted by Tamas Soltesz <ta...@codeandsoda.hu>.
Hi,
Thank you for the reply. When not providing the --master local[num] parameter,
it submits the train to the spark cluster. (see the MASTER env. variable in
the docker compose file.)
Now it seems like the exception is because the connection to the
elasticsearch is lost. Currently I am testing this idea (by not using the
elasticsearch at all).
BR,
Tamas Soltesz
On Tue, May 7, 2019 at 8:08 AM takako shimamoto <ch...@gmail.com>
wrote:
> Hi Tamas,
>
> Have you thought about submitting a training task to Spark Cluster?
>
> pio-docker train -- --master spark://spark-master:7077
>
> See:
> https://github.com/apache/predictionio/tree/develop/docker#run-with-spark-cluster
>
> On Fri, May 3, 2019 at 11:50 PM Tamas Soltesz
> <ta...@codeandsoda.hu> wrote:
> >
> > Hello guys,
> >
> > I am not sure about the process of asking questions here, so please be
> gentle with me :)
> >
> > I've asked my question on SO before I found this mailing list, would you
> please take a look there?
> >
> https://stackoverflow.com/questions/55967253/pio-train-fails-with-ioexception-connection-reset-by-peer
> >
> > Issue in short: I am running pio in docker swarm, and pio train fails
> with IOException: Connection reset by peer _after_ calculating and saving
> the model. I don't know how to find out what went wrong as I can't see any
> suspicious in any of the logs.
> >
> > If posting SO links here violates any rule here, please let me know. I
> can post my question here as well separately if needed.
> >
> > Thank you all,
> > Tamas Soltesz
>
Re: PredictionIO in docker - pio train fails
Posted by takako shimamoto <ch...@gmail.com>.
Hi Tamas,
Have you thought about submitting a training task to Spark Cluster?
pio-docker train -- --master spark://spark-master:7077
See: https://github.com/apache/predictionio/tree/develop/docker#run-with-spark-cluster
On Fri, May 3, 2019 at 11:50 PM Tamas Soltesz
<ta...@codeandsoda.hu> wrote:
>
> Hello guys,
>
> I am not sure about the process of asking questions here, so please be gentle with me :)
>
> I've asked my question on SO before I found this mailing list, would you please take a look there?
> https://stackoverflow.com/questions/55967253/pio-train-fails-with-ioexception-connection-reset-by-peer
>
> Issue in short: I am running pio in docker swarm, and pio train fails with IOException: Connection reset by peer _after_ calculating and saving the model. I don't know how to find out what went wrong as I can't see any suspicious in any of the logs.
>
> If posting SO links here violates any rule here, please let me know. I can post my question here as well separately if needed.
>
> Thank you all,
> Tamas Soltesz