MOO-cows Mailing List Archive

[Prev][Next][Index][Thread]

Re: [SERVER] memory leaks or MY MOO IS 8 TIMES BIGGER THAN IT USUALLY IS



At 05:01 PM 6/11/96 -0500, Diana L Harris <dlh5226@erc.jscc.cc.tn.us> wrote:
>I think I've noticed that almost all of the people with increased process 
>size have either used FUP or the regexp heavily. (Most being the former) As
>said by (sorry, forget the name) creator of FUP, FUP uses regexp and has some
>leaks...He said it'd be several weeks before a fixed version of FUP was
>released. (Please correct if I'm wrong.)

I'm well aware of this.  E_MOO has been using FUP since day 1; okay,
actually since like week 1.  All help is disk based, all web hits are disk
based, and queries are constantly made to the file system from the MOO via FUP.

Also, because of the in-db parser on E_MOO, regexps are commonly used on
each command that is entered.  Admittedly, this might not be as much of a
problem because the regexps that are used are relatively simple and the
regexp cache is set to 50 (as of when I compiled p5, it was the standard 20
under all previous versions)

E_MOO does exactly the same things that everyone is saying causes memory
leaks, yet I am not seeing any memory leaks coming from E_MOO.  This problem
is not track-down-able because it's not consistant and reproducable.  All we
have to go by is a list of regular expressions from Kipp (which we had to
beat out of him :) ) and Amy Bruckman's stack traces (which might turn out
to be an unrelated problem).  Everyone else has just said they are using FUP
or regexp, but fail to say what they are doing with FUP and what regexp they
are using.  Even by reading the changelog, you'll see that in the past, some
memory leaks have been caused by some very specific regular expressions.  We
need to know what those regular expressions are.

>> Is this memory leak thing just a bandwagon effect? Or perhaps people are
>> actually taking a look at the process lists for once and realizing how much
>> memory a MOO normally takes up. :)
>> 
>Nah, when you've seen you're MOO using 20% of the system's mem, and 
>suddenly it's up to 30% and still growing... then when ya restart it goes 
>back to normal...sounds like a leak to me.

It's the definition of 'normal' that I'm questioning.  
It's the use of 'and still growing' that I'm questioning.  No time frame has
been given in this case, and since ...
In no case should you expect the process size to remain the same as it was
when you started it up.  I am also questioning the experience of some of the
people who say that their MOO is growing out of control, especially when the
entire post consists of 'XMOO normally uses 9megs of memory, but I just
checked, and it's upto 15megs.  I restarted and it went back down to 9megs'.  

I guess what I'm trying to do is ask people to post the specifics about
their database along with the problem.  What exactly are you doing with FUP?
How complex are your regular expressions and what are they?  Are you doing
anything weird with binary strings?  Does your database contain heavy use of
exception handling?  What is in your extesions.c file?  These are the things
that we need to know.

Andy...
  (... join the "No More 'ME TOO' Posts" league).




Home | Subject Index | Thread Index