Experimental Physics and
Industrial Control System

Stephen Molloy <[email protected]> · Mon, 4 Jul 2016 21:09:42 +0000

Thanks Ralph,
Your comment on UDP & TCP was the clue I needed.  I had opened 5064 &
5065, but only for TCP connections.  Opening them up for UDP as well
allowed external connections.  Using the following command to run the
docker container did the trick,
docker run -i -t ‹rm -p 5064:5064 -p 5064:5064/udp -p 5065:5065 -p
5065:5065/udp sdmolloy/ubuntu-playground:latest /bin/bash

I¹m getting complaints about identical PV¹s being found on multiple
servers, but I think this is a side effect of running on a VM.

Thanks for your help,

Steve

On 04/07/16 17:02, "Ralph Lange" <[email protected]> wrote:

>Hi Steve,
>
>A CA client uses ephemeral ports as originating port numbers, correct.
>Ephemeral port numbers are indeed random, and cannot be predicted
>reliably.
>
>However, if your container contains an IOC (CA server), opening ports
>5064&5065 (in the default setup) for both TCP and UDP should allow any
>client outside the container to connect. (Not much different to a
>container with a web service published at port 80.)
>The CA client in the container would still use ephemeral port numbers
>for outgoing connections - do you need to publish those?
>
>Cheers,
>~Ralph
>
>
>On 04/07/2016 16:43, Stephen Molloy wrote:
>> I¹m trying to build a Docker container for quick set-up and tear-down
>> of EPICS instances, and I¹ve run into a problem.  A CA client outside
>> the container, but on the same host, cannot see any PV¹s served from
>> inside the container.
>> Looking at the network traffic with Wireshark, I can see that the
>> server uses ports 5064 & 5065 as expected, but the ports used by the
>> client seem to be relatively random.  If I could predict them, or
>> (even better) force them to be specific ports, then I could expose
>> those ports in the Docker container and (hopefully) the client and
>> server could communicate properly.
>>
>> Or perhaps this isn¹t the problem, and there is some other network
>> issue getting in my way.
>> (I¹ve successfully run several EPICS installations on this network and
>> from this computer, so I am confident that this is an issue with the
>> Docker set-up.)
>>
>> Hopefully I¹m on the right track with my thinking, and hopefully
>> someone can let me know how to proceed from here?
>>
>> In case you¹re interested, I¹m documenting my work on my blog ‹
>> http://www.smolloy.com/2016/07/easy-epics-implementation-with-docker/
>>
>> Thanks (again) for your time,
>>
>> Steve
>>
>

Experimental Physics and Industrial Control System

Experimental Physics and
Industrial Control System