> From: [EMAIL PROTECTED]
> 
> So if port.status is not a good option for determining what 
> state a process is in, what do people turn to?

Oh, I still use port.status, including LAYER.STACK, but only from tcl to
research a specific issue as it is happening.
I haven't tried the mass port.status-layer.stack data collection effort,
and I've never seen that problem again.

Whate else do people turn to?  
SYSTEM(9001) is a little different tool,  but at least tangential to the
current discussion.
It will give you the call stack of the current process, similar to what
you see with RAID "T" command or PORT.STATUS LAYER.STACK.   So I wrote a
subroutine that captures that plus some message you can send it and
saves it away for future examination.  Suppose I know a certain
subroutine occasionally throws errors.  Maybe it writes bad data or
uv/errlog occasionally shows an input argument unassigned.  That
subroutine gets called from mutliple places, varying execution levels
layers deep.  You don't know how it gets to that point.  I add this
debug logic to the suspect subroutine:
  
   IF [condition] THEN CALL *UVDBGLOGGER( control.parameter, [additional
message] )

Then I can examine the log and find out who, when, and the call
structure.  I have thought about having uvdbglogger send email, but so
far haven't.   I do have a periodic report that gets run and emailed.

SYSTEM(9001) is documented in this list's archives.
I'll give you the UVDBGLOGGER program to whoever wants it.

Chuck Stevenson
-------
u2-users mailing list
[email protected]
To unsubscribe please visit http://listserver.u2ug.org/

Reply via email to