Experimental Physics and
Industrial Control System

Matthias Clausen DESY -MKV2/KRYK- <[email protected]> · Mon, 26 Feb 1996 8:47:10 +0100 (MET)

>I think that this is a good idea.
good starting pont

>o I assume that the default would be to log writes to the log server? Hooks
>would be provided so that a site could choose to replace this default
>(and reroute the information to another destination).
agreed

>o Note that there will often be time periods when there are many writes
>(ie someone moves a valuator etc). In practice we may decide that only the
>last write in a given time period should be sent to the log. Does this imply
>that we will need a configuration parameter that selects a write log period?

configuration parameter either in the startup-script or through a special record or through Jim 
Kowalkowski's tool would be a right approach

>o Will we need to enable this feature only for particular records/fields?

all writes to any field would avoid additional configuration overhead

>o Since this would be logging significant changes it sounds like possible
>infrastructure for generalized backup and restore could be lurking here?

this would be a real positive side effect.
in the log-server one could configure that write to special fields will also go to a SQL-server 
which will write 'permanent' changes to our (future) ORACLE database from where we create the 
EPICS databases. This is one of our next projects and the logging service of the CA-server of 
course would cover a lot of necessary programming efford to create this functionality.

>o Does this functionality belong in the ca server or should it be implemented
>by each server tool? My initial thought is that implementing this in the server
>lib will minimizes lines of code. All configuration information would
>of course be obtained from the server tool.

I agree with Chip to keep this 'hidden' in the CA-server so noone can fool it.

Matthias

Experimental Physics and Industrial Control System

Experimental Physics and
Industrial Control System