On Friday, 06/13/2008 at 10:44 EDT, Colin Allinson 
<[EMAIL PROTECTED]> wrote:
> Today we were having occasional timeouts from RACF on one of the systems 
and I 
> could see that RACFVM was doing a steady 17.6 IO per second on this 
system. 
> 
> There is no activity on RACF on the other 3 systems and there are no 
DASD 
> reserves in place anywhere. 
> 
> I cannot see any reason for the activity (nothing in the console) and it 

> resumes even after I recycle RACF. 
> 
> I have looked for a looping user and cannot see anything chewing up load 
nor 
> any stream of RACF failures on the Op console.
> 
> It was causing the occasional timeout earlier when the systems were busy 
but 
> now it does not seem to be doing any harm. 
> 
> Does anyone have any other ideas where to look for what is causing this 
IO.

Where is the I/O going?  SMF log?  Database?  In general, RACF does 
nothing by himself.  #CP TRACE EXT RUN on the suspect RACFVM server will 
tell you if its run amok or if something is driving the horses.

The external interrupts can come from CP on the *RPI interface on from 
guests like LDAPSRV, FTPSERVE, and REXECD (issuing RACROUTEs).

QUERY IUCV RACFVM will give you the path ids.  If you've got QIUCV, I 
think it will tell you where the external interrupt buffer is in the 
RACFVM server.  My version does this:

QIUCV EXEC 1.2 (C) IBM Corporation, 1989, 2008 

User RACFVM has 4 out of a possible 300 connections. 
  - External interrupt buffer is located at 00E84D08 
  - Control interrupt buffer not defined 
  - 1208 messages received, 1209 sent, 0 failed 


But you can also go to each user listed in QUERY IUCV and do a TRACE IUCV 
on those users.

Alan Altmark
z/VM Development
IBM Endicott

Reply via email to