Re: [External] Trouble with pva:// on Phoebus with often-restarted IOCs

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

On 8/21/23 09:05, Leblanc, Gregory via Tech-talk wrote:
> Can I force it to listen on 5076 instead of falling back to another port?  It seems to be falling back to the dynamically assigned port number excessively.  I checked to see if anything was bound to 5076 with lsof |grep 5076 and ss -tulpn |grep 5076, but it still reported a dynamically assigned port.

Note the distinction between UDP and TCP ports.  All PVA peers (clients and servers)
will by default listen on UDP/5076.  (sharing a port is workable for connection-less
UDP, at least for receiving broadcasts)

The fallback to a dynamically assigned port number is for the TCP port as this can't be
shared between PVA servers.  You can override the default for each server/IOC, to
create a static mapping to match a firewall configuration.  eg.

> export EPICS_PVAS_SERVER_PORT=15075


> I don't think this is the whole story, because on my latest invocation of the IOC I did NOT get a dynamic port, but phoebus still told me that the PVs were disconnected, and it tried to look at a stale port number to communicate with the IOC.  Eventually I seem to be able to get Phoebus to switch back to the default ports, but so far it's just a process of trial and error to find something that works.
>       Greg
>
> > -----Original Message-----
> > From: Tech-talk <tech-talk-bounces at aps.anl.gov> On Behalf Of Leblanc, Gregory
> > via Tech-talk
> > Sent: Monday, August 21, 2023 10:26 AM
> > To: tech-talk at aps.anl.gov
> > Subject: [External] Trouble with pva:// on Phoebus with often-restarted IOCs
> >
> > Use caution with links and attachments.
> >
> > Hi folks,
> >
> > I'm working on developing an HMI here at the same time that I'm developing the
> > .db files.   I'm using the Phoebus nightly builds on a windows machine, and epics-
> > base checked out from github, compiled on CentOS 9 Stream.  I'm also trying to
> > use pva:// for all the PVs in Phoebus .  My typical workflow is to add a new record
> > to the .db file, stop the IOC, make distclean, make, and restart with st.cmd.
> > Next, I'll add a new widget to the Phoebus interface.  The first time I launch the
> > Phoebus display for the day, everything seems to work OK.  On all subsequent
> > times, the PVs show up in a disconnected state.  If I switch to ca://, I don't see this
> > issue.  I think this is somehow related to the dynamically assigned TCP port
> > numbers, but I'm unclear on what changes between the first launch and all
> > subsequent launches.  In my IOC output, I see:
> >
> > First launch:
> > 2023-08-21T09:13:00.122 Using dynamically assigned TCP port 38547.
> >
> > Next launch:
> > 2023-08-21T09:13:46.475 Using dynamically assigned TCP port 34715.
> >
> > And in the Error Log on Phoebus:
> > 2023-08-21 10:21:46 WARNING [org.epics.pva] Cannot connect to TCP
> > /10.0.0.241:34715
> >
> > I have confirmed that indeed I cannot telnet to the second port number listed
> > when things aren't working, although I haven't tried the telnet test when things
> > ARE working, but I'll try to do that today.  I've opened ports 5076/udp, 5075/tcp,
> > 5064 and 5065 both tcp and udp.  Any thoughts on where to go next?
> >       Greg
> >
> > --
> > Gregory Leblanc
> > Accelerator Engineer
> > Edwards Accelerator Lab - Ohio University
> > 123 University Terrace
> > Athens, OH 45701 USA
> > leblanc at ohio.edu
> > M: (401) 52-OUAL1 or (401) 526-8251