Experimental Physics and
Industrial Control System

"Kasemir, Kay (ky9)" <[email protected]> · Thu, 11 Sep 2008 09:13:29 -0400

Hi:

Is anybody using CAJ, the pure Java Channel Access implementation for Java,
in a production environment?

Which version of EPICS base?
Which version of CAJ?
With access security enabled on the IOC?
"Static" access security,
or acc. sec. rules which use CALC, or get re-loaded at runtime?

At the SNS, we still only use CAJ outside of the controls network,
because in the past it looked like CAJ could
result in CA Server/Access Security lockups on IOCs.
We were never able to reproduce it on demand,
but whenever an IOC's CA server was stuck in a call
to the access security library, at the worst possible time,
investigation revealed that somebody had accessed that IOC with CAJ.

Maybe that's been fixed in R3.14.9 and later versions of CAJ.
For us, JCA with R3.14.8.2 are just fine on the controls network,
and since we can't reproduce the lockup on demand
and thus are unable to check if it's fixed,
we're shying away from CAJ on the controls network.
On the office network, going through a CA gateway,
CAJ works very well and has obvious advantages for Java apps.

Thanks,
-Kay

[email protected] on 9/11/08 08:51 :

> Hi Daron,
> 
> Well, that document was created before cosylab posted CAJ version 1.1.5, and
> we have not done extensive testing on it to verify whether it address all of
> the known issues.
> 
> The serious issues associated with CAJ were:
> - could cause IOCs to crash
> - failed to connect to PVs when attempting to connect to a large number of PVs
> - poor PV connection performance
> 
> My testing of CAJ version 1.1.5 shows that it addresses the latter two issues
> (PV connection issues). However, I don't know the status of the most serious
> issue (IOC crashes).
> 
> At SNS we continue to use JCA/JNI exclusively for operations, but we sometimes
> use CAJ when connecting through the CA gateway since there are no such IOC
> crash issues in this context.
> 
> best regards,
> tom
> 
> 
> On Sep 10, 2008, at 5:07 PM, Daron Chabot wrote:
>> 
>> Hi Thomas,
>>  
>>  I was just looking at your slides, "Tour of XAL", and on the third-last
>> slide (pg 36) you note, "...known, serious issues with JCA/CAJ...", and
>> recommend instead the JCA/JNI combination for production use.
>>  
>>  Could you please elaborate on the problem(s) you hinted at ? I am exploring
>> JCA/CAJ and this information could be very helpful.
>>  
>>  Thanks.
>>  
>>  
>>  -- dc 
>>  
>>  
> 
> 

Experimental Physics and Industrial Control System

Experimental Physics and
Industrial Control System