Re: Ring buffer full error?

[Marty Kraimer <mrk@aps.anl.gov>]

Philip Taylor wrote:
> 
> When our system (the Gemini Telescope Control System,
> running EPICS 3.12.2) has been up for a while (usually a few
> hours) it starts giving out a unending stream of messages saying :
> 
>  interrupt: callbackRequest ring buffer full

> Its not clear whether this is actually causing (or
> associated with) any real problems. The system appears
> to continue running normally while the messages appear.
> 
> Does anyone have any suggestion as to what these
> messages might be caused by?
> 
> Philip
> 
> ------------------------------------------------------------
> Philip Taylor, Royal Greenwich Observatory,
> Madingley Road, CAMBRIDGE. CB3 0EZ. UK
> Email : pbt@ast.cam.ac.uk      Phone : (UK+) (0)1223-374710
> ------------------------------------------------------------

This is often caused by one of the general purpose callback tasks being
suspended. Are you certain you are on 3.12.2? I looked and in 3.12.2 the
messages is only issued when 1) the ring buffer is full and  2) a
callback was actually processed since the last time the message was
issued. The other possibility is that you are have very large databases
which use callbacks. In this case the queue size can be changed via the
following command (in the st.cmd file BEFORE iocInit)

callbackSetQueueSize(size)

The default size is 2000.

If, however, this only starts happening after several hours this may not
be the cause.

callbackRequest is the routine issuing the message. It is called by at
least the following:

1) asynchronous record completion.
2) I/O interrupt routines

Any other code is also allowed to use the call. One of the general
purpose callback tasks handles the callback.

Hope this helps.

Marty Kraimer