At 22/05/2000 09:28:00, you wrote:
>This is definitely a problem in your code. I routinely run tclhttpd for
>months without restarting or measurable growth in footprint.
Hello Ted.
Thank you very much for replying! I am definitely no expert at Tcl programming.
I try to get as much docs as I find, but still, sometimes a few doubts
remain:
- all code _is_ pure Tcl.
- if the problem are global definitions, why does the size at times actually
get smaller? I've one log record where the size abruptly goes to 0 for all
field in /proc/xx/statm, except total size. (without any action of mine!)
- why doesn't tclsh respond to kill -3 after a period of running? I'd like to
obtain a core dump (suggested by Brent) to try and analyze what's in memory.
I did recompile tclsh with debugging enabled.
Again, thanks!
John
>
>If your code is all in TCL, then examination of the code is the only
>solution. You could try getting rid of as many global variables as possible
>since that might help with your cleanup efforts.
>
>If your code is in C interfaced to TCL, then tools like purify may help you
>find your problem.
>
>-----Original Message-----
>From: John Coppens [mailto:[EMAIL PROTECTED]]
>Sent: Sunday, May 21, 2000 5:00 PM
>To: [EMAIL PROTECTED]
>Subject: Suggestions for memory leak problem?
>
>
>Hi all...
>
>I'd appreciate suggestions re this problem. Our University's www server,
>running tclhttpd 3.0.1, has a strange behaviour: it seems to occupy
>gradually more and more memory, starting at some 5% of system memory,
>and going to 20%. (See listing at the end of this mail) After that,
>somewhere, it seems to hang (occupying 99% of CPU, not replying to
>requests).
>
>After a while, following a suggestion of Brent, I tried to kill -QUIT the
>tclsh to obtain a core dump, but it does not seem to respond (kill -9
>works).
>
>The listing that follows is a cron script that dumps /proc/96/statm. I
>cannot
>find an exact definition of each field (pointers?), but even the first
>(total
>size) increases. And sometimes decreases.
>
>Can this be caused by a programming error of mine? I have the impression
>this
>is caused by some kind of buffer that doesn't get cleaned up...
>
>Thanks in advance!
>
>John
>
>
>Thu May 18 13:49:02 ART 2000
>945 927 314 2 0 925 613
>Thu May 18 13:50:01 ART 2000
>945 927 314 2 0 925 613
>Thu May 18 13:51:01 ART 2000
>945 927 314 2 0 925 613
>Thu May 18 13:52:01 ART 2000
>945 927 314 2 0 925 613
>Thu May 18 13:53:01 ART 2000
>933 915 314 2 0 913 601
>Thu May 18 13:54:01 ART 2000
>933 915 314 2 0 913 601
>Thu May 18 13:55:01 ART 2000
>947 929 314 2 0 927 615
>Thu May 18 13:56:01 ART 2000
>947 929 314 2 0 927 615
>Thu May 18 13:57:01 ART 2000
>955 937 314 2 0 935 623
>Thu May 18 13:58:01 ART 2000
>955 937 314 2 0 935 623
>Thu May 18 13:59:01 ART 2000
>955 937 314 2 0 935 623
>Thu May 18 14:00:01 ART 2000
>955 937 314 2 0 935 623
>Thu May 18 14:01:01 ART 2000
>955 937 314 2 0 935 623
>Thu May 18 14:02:01 ART 2000
>968 950 314 2 0 948 636
>Thu May 18 14:03:01 ART 2000
>968 950 314 2 0 948 636
>Thu May 18 14:04:01 ART 2000
>969 951 314 2 0 949 637
>Thu May 18 14:05:01 ART 2000
>993 975 314 2 0 973 661
>Thu May 18 14:06:01 ART 2000
>993 975 314 2 0 973 661
>Thu May 18 14:07:01 ART 2000
>993 975 314 2 0 973 661
>Thu May 18 14:08:01 ART 2000
>997 979 314 2 0 977 665
>Thu May 18 14:09:01 ART 2000
>1004 986 314 2 0 984 672
>Thu May 18 14:10:01 ART 2000
>1005 987 314 2 0 985 673
>Thu May 18 14:11:01 ART 2000
>1011 993 314 2 0 991 679
>Thu May 18 14:12:01 ART 2000
>1011 993 314 2 0 991 679
>Thu May 18 14:13:01 ART 2000
>1021 1003 314 2 0 1001 689
>----------------------------------------------------------------------------
>John Coppens
>Schools for Electrical/Electronics and Systems Engineering
>Faculty of Engineering
>Catholic University at C�rdoba, Argentina hamradio: lw3haz/on6jc
>mailto:[EMAIL PROTECTED] http://www.uccor.edu.ar/~jcoppens
>----------------------------------------------------------------------------
>
----------------------------------------------------------------------------
John Coppens
Schools for Electrical/Electronics and Systems Engineering
Faculty of Engineering
Catholic University at C�rdoba, Argentina hamradio: lw3haz/on6jc
mailto:[EMAIL PROTECTED] http://www.uccor.edu.ar/~jcoppens
----------------------------------------------------------------------------
