[tpm] Log4Perl

Fulko Hew fulko.hew at gmail.com
Tue Jan 29 06:37:08 PST 2013


On Sun, Jan 27, 2013 at 12:38 PM, Alex Beamish <talexb at gmail.com> wrote:

> Hi Tom,
>
> It is possible to change the Log4perl logging level without making source
> code changes -- you just adjust a value in the config file.
>

What I am trying to get our designers to add, is not just the ability to
change logging levels
without making source code changes, but to be able to change them without a
restart.

In our 'big' network, I'm trying to get people to standardize on SNMP, and
it allows viewing
_and_ changing things dynamically, (if the apps/devices support it).  In
production, nobody
wants to restart anything (unless its Windows in which case (sadly) the
standard answer to
anything is 'reboot'.), so the ability to alter the logging level without
restarting is very
beneficial.

Of course, it makes sense that making even that change on a Production
> machine would require permission. Changing the logging level would have a
> non-zero impact on the performance of the application, as well as the rate
> at which it consumed disk space (assuming you're logging to a disk file --
> there are lots of other ways that events can be logged, of course).
>

My anecdote is a system (I didn't write it) that when they turned on their
logging, it brought
the system to its knees because it was soo busy logging stuff across the
network to disks
that the system no longer had enough time left, nor network bandwidth, nor
disk bandwidth
to actually accomplish what it was trying to debug log.

... They quickly turned the logging off.

The moral of the story was... since they refused to build in the proper
run-time diagnostics
they had to resort to the 'detailed programmers debugging log file'...
which killed them
instead of helping them.  They should have done the job right... the first
time.


> Yes, everyone *wants* to write their own logging system -- of course, the
> reasoning is that a) existing solutions don't meet their needs, b) anyway
> they don't have time to learn someone elses's system, c) they won't be able
> to get permission to install someone else's code and d) it's just going to
> be something simple. These are all political reasons, and not technical
> reasons, and probably flow from a) programmer laziness, b) programmer
> hubris and c) edicts from PHBs who last wrote software when it was
> assembler for IBM mainframes ("in my day ..").
>

The usual rebuttal I get is: "its faster to invent and write my own 'xxx'
than it is to
learn they other guy's API, and get it integrated and working".
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mail.pm.org/pipermail/toronto-pm/attachments/20130129/26fc04c1/attachment.html>


More information about the toronto-pm mailing list