Experimental Physics and
Industrial Control System

Jeff Hill <[email protected]> · Wed, 04 Sep 1996 11:01:16 -0600

Gabor Csuka wrote:
> 
> Dear All !
> 
> We want to use the IOC log facility and during the test phase
> we decided to make some modification. Before we do this, ask
> YOU for comment.
> 
> regards,
> 
> Gabor CSUKA
> DESY, Germany
> 
> The proposal:
> 
> >
> > About IOCLoger:
> >
> >
> > Introduce:
> >
> >    The management for ioclog message is client/server mode. A server
> > can connect several different clients (ioc(s)),The Logging System
> > traps all logMsg calls and sends a copy to a Unix file .
> 
> Thats OK. But we want to use separate log files one per IOC.

This is currently possible. To do this you must run one log 
server deamon for each IOC. If you run more than one log server
deamon on the same host then of course each of them will need 
a unique port number via EPICS_IOC_LOG_PORT.

> 
> > The IOC log client can be disabled ,issuing the command
> >                                 iocLogDisable=1
> 
> According a code: It doesn't work, while the iocLogDisable
> variable is declared as 'LOCAL' not a 'GLOBAL' one. (As I see the code
> of R3.13, this bug is not there.)

As I recall this and several other log client related problems
were also fixed in a 3.12 bug fix release.

> 
> But if I insert this code before iocInit I have no chence to restart
> it with useing iocLogDisable=0 commmand. I prefeare 4 commands instead
> of one variable set:
> 
> iocLogClientStart - starts or restarts the ioclogclient if dosn't run
> iocLogClientStop  - stop (kill) the iocLogClient
> iocLogEnable      - disable the sending of iocLog messages to the Server.
> iocLogDisable     - enable sending the log messages
> 

Note that the log client auto reconnects if the log host is rebooted.
Given that it does reconnect then do we need the start/stop
functionality? Do you need start/stop so that the log client
can be forced to attach to a different server? I think that
the enbale/disable functionality is answered by the two
states of the boolean iocLogDisable variable?

> The envinroment variables:
> >         EPICS_IOC_LOG_FILE_NAME=/applic/log/iocLog.text
> 
> usenig as a direcory name, and separate files for IOC-s

If we did this we would need to handle situations where the 
IOC's IP address could not be converted to an ASCII host name. 
Some sites will continue to prefer to keep several IOC's in 
one file?

> >         EPICS_IOC_LOG_FILE_LIMIT=1000000
> 
> keepeing it as a limit /ioc-s.

If we did this we would need a mechanism to configure the
per IOC file name and file limit. Would we use an env
variable or some ascii config file? 

We can satisfy the above request by running multiple servers.
Is the above approach significantly better. Is the added
complexity in the log server justified (it might be if
there are many IOCs at a particular site)?

> 
> >
> > Once the file is written fully, it will be  overwrite by new messages
> > from the beginning of the file.
> 
> Ok. but needed a possibility to inform the system manager about...
> 

It would be easy to provide an option where the server creates
a new file with the current date appended to the file name
when it gets to the end of the current file?

I also recall that DESY and KEK have requested that the
option of logging every write (put) operation that goes
through the server (including the client's login name/host name).

I have many other items in my queue at this point. If DESY
(or some other site) arrives at a set of changes which meets
the approval of this list and they are also interested in
performing the work then we will certainly integrate the result 
into a future release.

Jeff

-- 

______________________________________________________________________
Jeffrey O. Hill			Internet	[email protected]
LANL MS H820			Voice		505 665 1831
Los Alamos, NM 87545 USA 	FAX		505 665 5107

Experimental Physics and Industrial Control System

Experimental Physics and
Industrial Control System