But I would like to know
if this is a familiar problem or that running CA Gateways on VMWare
is a bad idea to begin with (I had big latency problems many years
ago, so I went to a hardware solution at that time).
Generally, I would say no to both questions.
For a Gateway, CA put() with callback requests are pretty hard to deal with correctly, especially if the target record is asynchronous and takes a while to process and multiple clients are trying to do this concurrently. Just to say that I am not surprised that this kind of operation is causing trouble.
I don't think we have seen this particular issue before; it is certainly not familiar. There are two obviously bad things under investigation: see Gateway issues
33 and
35. Hugo's idea would verify if what you see is related to the latter of those.
I honestly can't remember running a Gateway on bare metal. At ITER, only some front end controllers are direct hardware installs, everything else is virtualized. There's definitely no conceptual reason why running a Gateway on a virtual machine would be a bad idea.
Of course, depending on your use case, there are performance and bandwidth limitations that could affect a Gateway, which is also true for virtual systems.
EPICS 3.14.12.4 is pretty old (released end of 2013).
There have been fixes and improvements in Channel Access since then (even small protocol changes), and while mix-and-match of EPICS versions usually works without issues, I would suggest keeping Gateways on a supported release of EPICS Base.
Cheers,
~Ralph
