Experimental Physics and
Industrial Control System

Jeff Hill <[email protected]> · Fri, 03 May 1996 10:26:13 -0600

William Lupton wrote:
> 
> Dear all,
> 
>   We are discussing how to handle simulation both when IOCs are present and
> when they are absent. For simulation at other sites we are writing a Unix
> CA server which emulates part of our VME database. For simulation at the
> telescope we are toying with the idea that we might run the Unix CA server
> all the time but that connections to it would only be used when the vxWorks
> CA server was not running.
> 
>   This raised the question of whether anyone has ever discussed implementing
> "fallback" CA servers which would automatically be used (perhaps only for
> gets?) if the "master" server failed. At the CA server level it would imply
> that the server had an associated priority level. At the CA client level,
> multiple responses to ca_search() would have to make use of the priority
> level to decide which server to connect to, and connection state change
> could cause server rollover.
> 

Tate - one of the industrial partners - has a version of EPICS which
supports dual redundant IOCs. If one IOC fails the other IOC detects it and
control of the IO is switched over to the properly functioning IOC.

I agree with Steven Beard that we would need to be careful that the
operator isnt mislead to think he is controlling the operational
system when in fact he is still attached to the simulation server.

I think the idea of a backup server that tracks the values in the
operational system is an interesting idea. The backup/simulation
server would need to receive proper notification when the master is ok.
Presumably the backup server would then disconnect its clients and the
clients would then reattach to the master.

Support for this would require changes to the CA protocol I suspect.

Jeff

Experimental Physics and Industrial Control System

Experimental Physics and
Industrial Control System