Experimental Physics and
Industrial Control System

Ralph Lange <[email protected]> · Tue, 06 Nov 2007 17:05:50 +0100

Hi -
(putting this on tech-talk instead of CCing a dozen people since it
might be of interest to a broader audience)
the Screen utility is what the original APS iocConsole (and SLAC's
version of it) is based on, and it's what BESSY is using so far.
IocConsole uses Screen's ability to allow multiple users parallel access
to the soft IOC console as well as Screen's logging mechanism.
BESSY has put Screen as the "soft IOC container" below Conserver, a free
client/server software that manages logging and parallel access
(including authorization and authentication) to all serial type
connections, thereby unifying the access to all IOCs, soft and VME based.

Therefore, we do not use the parallel access and logging mechanisms of
Screen: Only one client (the Conserver) is logged into the Screen
session running the IOC all the time. (I put a wiki page on the EPICS
web about this.)
Stephanie Allison (SLAC) has seen softIOCs hang when people use the
history scrolling mechanism of Screen.

We found that Screen puts ugly vt100 control characters into Conserver's
log files.
The SNS is using a small wrapper (called procServ) around their soft
IOCs that provides telnet access to stdin and stdout of the IOC as well
as file logging. It also restarts the softIOC when it terminates to
allow rebooting softIOCs, and writes a PID file. Blowing this up to
allow parallel access to the console is still on their to-do-list.
To me, the best solution at the moment sounds like running the soft IOC
under procServ (just as a minimal IOC shell) and Conserver on top (to
provide shared access, logging, integration of VME and soft IOCs).

To forward the log content to Conserver, I'm right now trying to have
procServ log into a named pipe, so that Conserver - when it reconnects -
still gets the complete history of a rebooting soft IOC. (Conserver will
telnet to the procServ port for access and ssh to cat the named pipe for
logging.)

I will also try to allow a certain exit code from the IOC keeping
procServ from restarting the IOC, so that orderly shutdown and reboot
will be available from the iocsh.
Corrections, encouragement, ideas anyone?
Stay tuned ....
Ralph

Dave Reid wrote:

Heinrich, Ralph--
I was digging around on Tech-Talk looking for some information about
starting up soft IOC's using scripts, and saw your notes about using
the SCREEN utility.
We are trying to use that here, but we've run into somewhat of an
interesting issue. I was looking at a running IOC this morning, and
when I tried to "detach", using the CTL+A+D, for some reason the "A"
part didn't take, and I ended up crashing the IOC. Obviously, this
doesn't go over very well.
Is using the SCREEN utility still the only way to do something like
this? Have you ever run into a problem like this with inadvertently
crashing an IOC?
Thanks for any input...
Dave Reid
University of Washington Medical Center

Experimental Physics and Industrial Control System

Experimental Physics and
Industrial Control System