The current developers guide (developers-guide.html) states that
"catch-and-wrap" should be logged as 'info' or 'error'. (of course, I
wrote that... and nobody veto'd :-)
The situation is that when we "wrap" an internal error as an AxisFault then
we do NOT want to deliver to the client... it's an internal problem that
doesn't need to be exposed (and the code needs work to adjust to this
philosophy). On the other hand, we must be able to go back to the logs and
research the problem (first failure problem resolution... in production we
don't always have the luxury of being able to go back and attempt to
recreate a problem).
Again, that our goal should be AXIS in a production, enterprise-ready,
environment... what looks nice on the screen to the developers during test
is not relevant. If it's causing problems in verifying operation of tests,
then we need to find another way to verify, NOT abandon robust coding
practices.
*******************************************
Richard A. Sitze [EMAIL PROTECTED]
CORBA Interoperability & WebServices
IBM WebSphere Development
R J Scheuerle
Jr/Austin/IBM@IB To: [EMAIL PROTECTED]
MUS cc:
Subject: RE: logging question: catch
(Exception) throw AxisFault
05/10/2002 12:03
PM
Please respond
to axis-dev
The development doc needs to be changed to reflect the consensus.
Here are some rules to get the discussion moving:
} catch (Exception e) {
// Log or not log ?
throw new AxisFault(...);
}
1) If the AxisFault is simply wrapping another exception, then nothing
should be logged (except perhaps debug).
2) If the AxisFault is not wrapping another exception, or if a
non-AxisFault message is thrown, at least an info message should be issued.
3) There will be any deviations to these rules. The deviations should be
clearly defined in a comment which indicates the reason for the deviation.
(This will prevent the code from being changed back and forth.)
Rich Scheuerle
XML & Web Services Development
512-838-5115 (IBM TL 678-5115)
Russell
Butek/Austin/IBM@IBMUS To: [EMAIL PROTECTED]
cc:
Subject: RE: logging
05/10/2002 10:58 AM question: catch (Exception) throw AxisFault
Please respond to
axis-dev
I can appreciate that exceptions should be logged, and they ARE still
logged if you turn on debug. The immediate problem (and there are lesser
ones in the build) is the multithread test case. When that test pounds the
server with 400 requests it WILL get some connection refused exceptions
(for me, at least, this is usually only a few, but I've had as many as
360). My test ignores these exceptions, but the build output is flooded
with them when log.info is used.
Russell Butek
[EMAIL PROTECTED]
Tom Jordahl <[EMAIL PROTECTED]> on 05/10/2002 10:47:55 AM
Please respond to [EMAIL PROTECTED]
To: "'[EMAIL PROTECTED]'" <[EMAIL PROTECTED]>
cc:
Subject: RE: logging question: catch (Exception) throw AxisFault
+1
See http://marc.theaimsgroup.com/?l=axis-dev&m=102028751728496&w=2
But Richard feels that all exceptions should be logged.
--
Tom Jordahl
-----Original Message-----
From: Russell Butek [mailto:[EMAIL PROTECTED]]
Sent: Friday, May 10, 2002 11:42 AM
To: [EMAIL PROTECTED]
Subject: logging question: catch (Exception) throw AxisFault
Various places in the code are of the form
} catch (Exception e) {
log.xxx(... e);
throw new AxisFault(...);
}
In some places xxx is debug, others it is info. I contend that it should
always be debug. When it's info, then the build's output contains stack
dumps that make reading the output a nuisance. (These outputs weren't
there in the past.)
Russell Butek
[EMAIL PROTECTED]
