Experimental Physics and
Industrial Control System

Jeff Hill <[email protected]> · Thu, 30 Oct 1997 16:13:03 -0700

Stephanie,

I fixed a CA reconnect problem occurring when IOCs boot faster than 
EPICS_CA_CONN_TMO
on Aug 4th of this year. Looking in CVS it appears that this fix is _not_ in 
3.13.beta11 or
any other release of EPICS 3.13. I suspect that this bug has been in EPICS for 
a long time,
but not discovered until now because the IOCs frequently reboot faster under 
3.13.

Here is the bug report that I should have placed on the mail list:

Symptom:
A CA client sometimes does not reconnect if the server (IOC) reboots faster 
than EPICS_CA_CONN_TMO
and no requests are made by the client after the server reboots.

A fix will show up in the next release after EPICS 3.13.beta11. The problem was 
discovered at APS
and I have not heard any reports of additional reconnect problems from them 
since the fix was installed.

Jeff

On Thursday, October 30, 1997 10:59 AM, [email protected] 
[SMTP:[email protected]] wrote:
> Hello -
>
> Should the beacon period be set less than the time that an IOC needs
> to boot and initialize?  If yes, how much less?  Is there a rule of
> thumb?  I'm trying to track down some CA client reconnect problems.
> I found that our VMS CA clients were not polling often enough -
> increasing the polling interval from 22(!) to 2 seconds has solved
> the reconnect problem for IOCs that take minutes to boot (we're still
> on beta4).  I still see the problem for the faster booting
> (~15 seconds) IOCs, particularly after a crate power-down and power-up.
> I wonder about our 15 second beacon period.
>
> Thanks for any insights,
>
> Stephanie Allison  [email protected]

Experimental Physics and Industrial Control System

Experimental Physics and
Industrial Control System