RE: problem of memory with tcptrace

From: zze-PELLEGRIN Jerome stagiaire FTRD/DMI/SOP (jerome.pellegrin@rd.francetelecom.fr)
Date: 02/22/01

  • Next message: Rick Jones: "Re: problem of memory with tcptrace" 

    Message-ID: <D5FD8AC43CFED3119A8E00508B6FECFE42D8C2@p-merle.rd.francetelecom.fr>
From: zze-PELLEGRIN Jerome  stagiaire FTRD/DMI/SOP <jerome.pellegrin@rd.francetelecom.fr>
Subject: RE: problem of memory with tcptrace 
Date: Thu, 22 Feb 2001 17:49:16 +0100

    Thanks for your response

    i have done what you said and it's a problem of swap, when there is only 60
    Mo of swap tcptrace stop

    in fact 1Go of am and 1Go of swap it's not enough for my file !!! :(
    i attach the file made by truss, in fact only the last 5000 lines

    i think i will try to add some swap file like you said

    thanks for your help

    jerome :)

    -----Message d'origine-----
    De : Shawn Ostermann [mailto:ostermann@cs.ohiou.edu]
    Envoye : jeudi 22 fevrier 2001 16:36
    A : zze-PELLEGRIN Jerome stagiaire FTRD/DMI/SOP
    Cc : tcptrace@tcptrace.org
    Objet : Re: problem of memory with tcptrace

    > I want to used tcptrace to analyse traffic of a web server
    > I have made trace with tcpdump,
    > the binaries have been zipped on the fly
    > the size of the zipped binaries are between 700 Mo and 1,6Go
    > but they seems to be to big because when i launch tcptrace , the error
    > "Ressource temporaly unavailable"
    > accurs.
    > I used a Sun ultrasparc2 with solaris2.6 an 1Go of ram
    >
    > Do you know a solution to compute my files whitout cut them in small part
    >
    > sorry if my question is stupid
    >
    > jerome :)

    It's not a stupid question at all. It isn't really a tcptrace error,
    it's coming from the operating system. You don't give quite enough
    information to know for sure, but I suspect that it's a memory
    allocation problem. You can get that from malloc(), although I'm a
    little confused because we have a malloc wrapper that traps this error
    and tries to explain it a little. Perhaps you're seeing that and
    didn't mention it. In any case, if that's the problem, then you're
    out of swap space (but may or may not be out of physical memory). A
    quick way to verify this is to run 'top' (freely available) and watch
    the amount of available swap space as the program runs, if it goes to
    0 right before the program exits, that's your problem. Adding
    temporary swap space on a solaris machine (and most others) is fairly
    simple. You would use a combination of mkfile(1M) to make a large
    file to swap into (maybe 1GB) and then swap(1M) to add that new file
    to the list of swap space. This is a little slower than a real swap
    partition, but much less work. Note that this requires root access.

    If I'm way off base here, please write back with some more details.
    I'd specifically like to see the output of:

      truss -oTRUSSFILE tcptrace -args file...

    then the file 'TRUSSFILE' will show which system call is failing and
    why. For those of you not on Solaris, a truss-like program can
    probably be found as 'strace', 'ptrace', or sometimes just 'trace'.

    --sdo
    ----------------------------------------------------------------------------
    To unsubscribe, send a message with body containing "unsubscribe tcptrace" to
    majordomo@tcptrace.org.

    This archive was generated by hypermail 2b30 : 02/22/01 EST