Groups > comp.lang.java.programmer > #4529

Re: analysis of java application logs

From	Martin Gregorie <martin@address-in-sig.invalid>
Newsgroups	comp.lang.java.programmer
Subject	Re: analysis of java application logs
Date	2011-05-24 12:29 +0000
Organization	UK Free Software Network
Message-ID	<irg8b1$ibj$1@localhost.localdomain> (permalink)
References	(1 earlier) <irdmdc$hfg$1@news.albasani.net> <ire4op$fgq$1@dont-email.me> <ireave$1m2$1@news.albasani.net> <iremt2$u3k$1@localhost.localdomain> <941ivgF1amU1@mid.individual.net>

Show all headers | View raw

On Tue, 24 May 2011 12:26:39 +0100, Nigel Wade wrote:

> On 23/05/11 23:25, Martin Gregorie wrote:
>> On Mon, 23 May 2011 15:02:23 -0400, Lew wrote:
>> 
>>> On 05/23/2011 01:16 PM, Daniele Futtorovic wrote:
>>>> Let me emphasize: IMHO debugging logs and logs for analysis are two
>>>> different things and should be kept strictly separated -- possibly
>>>> logged to a different target respectively.
>>>
>>> That last is rather a brilliant idea, to use different targets.
>>> Heretofore I've espoused that logs are primarily an operations tool,
>>> not a debugging tool, although in service of the former they
>>> inevitably and inherently must support the former.  The problem I've
>>> always seen is that logging statements are left up to the programmer,
>>> and not specified for the project.
>>>
>> I tend to use at least two logging streams: debugging and operational.
>> I leave debugging statements in production code: its normally off (of
>> course) but can be turned on if needed.
> 
> There is one caveat to leaving debug logging in production code; it may
> affect performance. Even with output disabled the string arguments are
> still constructed, unless they are constant. If logging is located in a
> tight loop, or critical section, it might become significant.
>
Fair point. However, my usual debugging statement takes the form:

if (debug > 0)
  debugger.trace(result + " = method(" + arg + ")");

Ugly, I know, but quite efficient, since when debugging is off even the 
cost of the method call cost is. I use an integer to control debugging 
rather than a boolean so I can control its volume: "java Application -dd" 
would be expected to provide more detailed debugging output than "java 
Application -d"

> In one particular application I profiled I found this to be eating up
> 80% of the CPU time. Admittedly, this was a rather special case, but
> it's still something which ought to be borne in mind if you leave debug
> logging code in production software.
>
Agreed. 

On occasion I've used a circular buffer to accumulate tracing information 
which gets only dumped if an exception occurs. This is great if you're 
chasing rarely occurring problems in a high volume, long running process 
since it obviates searching through megabytes of tracing info to find 
20-30 lines of relevant tracing. You'd expect this too to be quite 
expensive to run, so I'd use the same mechanism outlined earlier to 
ensure that tracing set off means that circular buffer is never filled 
rather than merely suppressing the buffer dump operation.

That said, I used to administer an OS (ICL's George 3) that continuously 
traced its internal operations to a fine-grained and a coarse-grained 
circular buffer which could be dumped after a crash and still managed to 
run fast enough to be a very usable OS.

-- 
martin@   | Martin Gregorie
gregorie. | Essex, UK-
org       |

Back to comp.lang.java.programmer | Previous | Next — Previous in thread | Next in thread | Find similar

Thread

analysis of java application logs Ulrich Scholz <d7@thispla.net> - 2011-05-23 00:50 -0700
  Re: analysis of java application logs Robert Klemme <shortcutter@googlemail.com> - 2011-05-23 02:20 -0700
  Re: analysis of java application logs jlp <jlp@jlp.com> - 2011-05-23 11:45 +0200
  Re: analysis of java application logs Lew <noone@lewscanon.com> - 2011-05-23 09:11 -0400
    Re: analysis of java application logs Daniele Futtorovic <da.futt.news@laposte-dot-net.invalid> - 2011-05-23 19:16 +0200
      Re: analysis of java application logs Lew <noone@lewscanon.com> - 2011-05-23 15:02 -0400
        Re: analysis of java application logs Daniele Futtorovic <da.futt.news@laposte-dot-net.invalid> - 2011-05-23 22:03 +0200
          Re: analysis of java application logs Michael Wojcik <mwojcik@newsguy.com> - 2011-05-26 14:43 -0400
        Re: analysis of java application logs Arved Sandstrom <asandstrom3minus1@eastlink.ca> - 2011-05-23 18:32 -0300
          Re: analysis of java application logs Ulrich Scholz <d7@thispla.net> - 2011-05-25 06:00 -0700
            Re: analysis of java application logs Arved Sandstrom <asandstrom3minus1@eastlink.ca> - 2011-05-25 19:04 -0300
        Re: analysis of java application logs Martin Gregorie <martin@address-in-sig.invalid> - 2011-05-23 22:25 +0000
          Re: analysis of java application logs Nigel Wade <nmw-news@ion.le.ac.uk> - 2011-05-24 12:26 +0100
            Re: analysis of java application logs Martin Gregorie <martin@address-in-sig.invalid> - 2011-05-24 12:29 +0000
              Re: analysis of java application logs Lew <noone@lewscanon.com> - 2011-05-24 08:49 -0400
                Re: analysis of java application logs Martin Gregorie <martin@address-in-sig.invalid> - 2011-05-24 14:37 +0000
                Re: analysis of java application logs Lew <noone@lewscanon.com> - 2011-05-24 12:26 -0400
                Re: analysis of java application logs Jim Gibson <jimsgibson@gmail.com> - 2011-05-24 11:00 -0700
                Re: analysis of java application logs Lew <noone@lewscanon.com> - 2011-05-24 14:35 -0400
                Re: analysis of java application logs Nigel Wade <nmw-news@ion.le.ac.uk> - 2011-05-25 09:53 +0100
            Re: analysis of java application logs Daniele Futtorovic <da.futt.news@laposte-dot-net.invalid> - 2011-05-24 19:12 +0200
  Re: analysis of java application logs Patricia Shanahan <pats@acm.org> - 2011-05-23 06:17 -0700
    Re: analysis of java application logs Robert Klemme <shortcutter@googlemail.com> - 2011-05-23 20:33 +0200
      Re: analysis of java application logs Martin Gregorie <martin@address-in-sig.invalid> - 2011-05-23 19:07 +0000
  Re: analysis of java application logs CncShipper <anon@nowhere.com> - 2011-05-23 14:56 +0000
    Re: analysis of java application logs Lew <noone@lewscanon.com> - 2011-05-23 11:43 -0400
      Re: analysis of java application logs jlp <jlp@jlp.com> - 2011-05-23 18:00 +0200
        Re: analysis of java application logs Lew <noone@lewscanon.com> - 2011-05-23 12:20 -0400
          Re: analysis of java application logs Daniele Futtorovic <da.futt.news@laposte-dot-net.invalid> - 2011-05-23 19:06 +0200
            Re: analysis of java application logs Robert Klemme <shortcutter@googlemail.com> - 2011-05-23 20:27 +0200
              Re: analysis of java application logs Daniele Futtorovic <da.futt.news@laposte-dot-net.invalid> - 2011-05-23 21:02 +0200
              Re: analysis of java application logs Lew <noone@lewscanon.com> - 2011-05-23 15:06 -0400
                Re: analysis of java application logs Robert Klemme <shortcutter@googlemail.com> - 2011-05-23 22:10 +0200
                Re: analysis of java application logs Lew <noone@lewscanon.com> - 2011-05-23 17:04 -0400
  Re: analysis of java application logs Tom Anderson <twic@urchin.earth.li> - 2011-05-24 15:04 +0100
    Re: analysis of java application logs Martin Gregorie <martin@address-in-sig.invalid> - 2011-05-24 14:50 +0000
      Re: analysis of java application logs Michael Wojcik <mwojcik@newsguy.com> - 2011-05-26 14:58 -0400
      Re: analysis of java application logs Lawrence D'Oliveiro <ldo@geek-central.gen.new_zealand> - 2011-05-30 16:23 +1200
        Re: analysis of java application logs Lew <noone@lewscanon.com> - 2011-05-30 01:08 -0400

csiph-web