Re: EDM freezes

[Brian Bevins <bevins@jlab.org>]

John Sinclair wrote:
> > > 2) There is a more subtle ongoing problem that (as far as I know) 
> > > has only been reported here at Jefferson Lab. When an EDM session 
> > > has been left running for a long (variable) time, the widgets will 
> > > stop updating despite the fact that the PV values are still 
> > > changing. There is no visual clue that anything is wrong, unless you 
> > > can see the same PV's in a different app or a separate EDM session. 
> > > Usually everything will start updating again if you open or close a 
> > > display, or even expose a covered display. This issue only started 
> > > when we moved from HP-UX to RedHat Enterprise Linux. It has not yet 
> > > been resolved.
>
> It would be interesting to know more about this. If you mouse over
> a display element like a text control and do a control-middle-mouse 
> click you will get something like the following in the message window.
>
> ------------------------------------------------
>
> Text Control (105,46): b:Vgc12-1Ig1:PressureR
>   Host: orib60.phy.ornl.gov:5064
>   Num times conected = 1
>   Num times disconected = 0
>   Num value changes = 1
>   Num references = 2
>   Time stamp = Wed Jan 21 17:11:00 2009
>
> ------------------------------------------------
>
> It would be nice to know if the various pvs that were no longer updating
> had any things in common like hostname, timestamp, or if any of the 
> numeric values looked fishy (negative values, etc.).
>
> John Sinclair

We've been trying to find patterns in the occurrences, but so far 
nothing has emerged. We cannot reliably reproduce the problem. I have 
never had it happen on my own workstation no matter how long I keep EDM 
running, and many operators never see it. Many user interactions (window 
open or close, expose events, sometimes clicking control widgets) will 
kick everything into working order again, so the data is sparse.

I do have reports that ctrl-middle clicking on "frozen" widgets multiple 
times does show that CA updates are being received and timestamps are 
increasing, but I have not been able to verify this myself. Based on 
reports, there is no consistency in the server hostname of the frozen 
widgets. They seem to freeze screen by screen, rather than ioc by ioc. 
The values displayed in the widgets never look fishy, and ctrl-middle 
click does not report the current value.

--Brian

--
Brian S. Bevins, PE
Computer Scientist / Mechanical Engineer
Thomas Jefferson National Accelerator Facility

    "Nothing in all the world is more dangerous than
     sincere ignorance and conscientious stupidity."
                                    --Martin Luther King Jr.