Re: p4p timeout on get() in broadcastless network

[Michael Davidsaver via Tech-talk <tech-talk at aps.anl.gov>]

On 1/23/26 9:40 AM, Ulrik Pedersen via Tech-talk wrote:

Hi Freddie,

Thanks for pitching in! This is interesting - it does appear to solve the problem!!! I had not even thought of that one because I don't run a name service... Thanks for the hint - I will study this a bit closer to understand why this works!

In brief.  With PVXS, every server is a TCP nameserver for itself.

https://epics-base.github.io/pvxs/netconfig.html

In the meantime, I have also found another workaround: if I set 127.0.0.1 in the front of my EPICS_PVA_ADDR_LIST then it works both for local host IOCs and other IOCs on the network. 

Recent PVXS can make use of the loopback for regular search traffic.

I'm still baffled why p4p behaves differently from phoebus and the EPICS base pva cli tools which works without setting these variables.

Three different implementations, which will almost inevitably have different behaviors.

Also, to make the situation even more complex.  With UDP unicast traffic, it makes a difference which PVA peer is started first (on Linux) or last (on Windows).  This is the peer which will handle the automagic forwarding of unicasts to all peers.

Before I could say more about what specifically is going on, I would first need more information about your virtual networking environment.

Cheers,

Ulrik

 

On Fri, 23 Jan 2026 at 17:13, Freddie Akeroyd - STFC UKRI <freddie.akeroyd at stfc.ac.uk> wrote:

Hi Ulrik,

 

Does setting

 

    EPICS_PVA_NAME_SERVERS=10.0.0.12

 

For the client help?

 

Regards,

 

Freddie

 

From: Tech-talk <tech-talk-bounces at aps.anl.gov> On Behalf Of Ulrik Pedersen via Tech-talk
Sent: 23 January 2026 16:44
To: Tech talk <tech-talk at aps.anl.gov>
Subject: p4p timeout on get() in broadcastless network

 

Hello,

 

I am having some trouble getting p4p working reliably in a cloud based network where UDP broadcasts are not possible. I am only using a single VM, running both the IOC and the p4p client on the same VM. The setup works well with Phoebus, archiver, and pvget/pvput/pvmonitor from EPICS base, just not with p4p...

 

I am running an EPICS IOC with PVXS and QSRV2 and just a few PVs. The IOC is running with the following EPICS environment to work around the fact that we can't use broadcast (the 10.0.0.12 is the VMs IP on a private network with archiver, etc.):

 

user02@uservm02:~$ env | grep -i pva | sort

EPICS_PVAS_AUTO_BEACON_ADDR_LIST=NO

EPICS_PVAS_BEACON_ADDR_LIST=10.0.0.12

EPICS_PVAS_BROADCAST_PORT=5076

EPICS_PVAS_SERVER_PORT=5075

EPICS_PVA_ADDR_LIST=10.0.0.12

EPICS_PVA_AUTO_ADDR_LIST=NO

EPICS_PVA_BROADCAST_PORT=5076

 

Then on for the p4p client, I am running with this environment:

 

(epics-training-env) user02@uservm02:~$ env | grep -i pva | sort

EPICS_PVAS_AUTO_BEACON_ADDR_LIST=NO

EPICS_PVAS_BEACON_ADDR_LIST=10.0.0.12

EPICS_PVA_ADDR_LIST=10.0.0.12

EPICS_PVA_AUTO_ADDR_LIST=NO

EPICS_PVA_BROADCAST_PORT=5076

 

(I know the PVAS variables are not needed on the client. I have been tinkering a lot with these today)

 

My python script with p4p looks like this:

 

(epics-training-env) user02@uservm02:~$ cat small.py 

from p4p.client.thread import Context

ctxt = Context('pva')

print(ctxt.get('user02:va:01:gauge:pressure'))

ctxt.close()

 

Sometimes the script will run successfully once or twice, but then it will mostly timeout with this error dump:

 

(epics-training-env) user02@uservm02:~$ python small.py 

Traceback (most recent call last):

  File "/opt/epics/python/.venv/lib/python3.12/site-packages/p4p/client/thread.py", line 264, in get

    value, i = done.get(timeout=timeout)

               ^^^^^^^^^^^^^^^^^^^^^^^^^

  File "/usr/lib/python3.12/queue.py", line 179, in get

    raise Empty

_queue.Empty

 

During handling of the above exception, another exception occurred:

 

Traceback (most recent call last):

  File "/home/user02/small.py", line 3, in <module>

    print(ctxt.get('user02:va:01:gauge:pressure'))

          ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^

  File "/opt/epics/python/.venv/lib/python3.12/site-packages/p4p/client/thread.py", line 268, in get

    raise TimeoutError()

TimeoutError

 

I am on an Ubuntu 24.04 VM. Using system python 3.12.3, p4p==4.2.2

 

Hope someone can point out my silly mistake here... 

 

Cheers,

Ulrik