using RHEL / CentOS / Fedora perl? - SpamAssassin

This is a discussion on using RHEL / CentOS / Fedora perl? - SpamAssassin ; have you seen this? http://blog.vipul.net/2008/08/24/red...hat-a-tragedy/ That bug in Red Hat perl will almost definitely slow down SpamAssassin, too, I would say. Can anyone verify? --j....

+ Reply to Thread
Results 1 to 9 of 9

Thread: using RHEL / CentOS / Fedora perl?

  1. using RHEL / CentOS / Fedora perl?

    have you seen this?

    http://blog.vipul.net/2008/08/24/red...hat-a-tragedy/

    That bug in Red Hat perl will almost definitely slow down SpamAssassin,
    too, I would say. Can anyone verify?

    --j.


  2. Re: using RHEL / CentOS / Fedora perl?


    On 8-Sep-08, at 6:38 AM, Justin Mason wrote:

    > have you seen this?
    >
    > http://blog.vipul.net/2008/08/24/red...hat-a-tragedy/
    >
    > That bug in Red Hat perl will almost definitely slow down
    > SpamAssassin,
    > too, I would say. Can anyone verify?
    >
    > --j.


    I know as far as centos goes, there are some patches out already by
    Karanbir Singh.

    http://www.karan.org/blog/


  3. Re: using RHEL / CentOS / Fedora perl?

    Justin Mason wrote:
    > have you seen this?
    >
    > http://blog.vipul.net/2008/08/24/red...hat-a-tragedy/
    >
    > That bug in Red Hat perl will almost definitely slow down SpamAssassin,
    > too, I would say. Can anyone verify?
    >
    > --j.


    I can verify that CentOS release 4.6 (Final) with perl, v5.8.5 built
    for i386-linux-thread-multi is NOT affected.

    time perl perlbug.pl
    .................................................. .
    real 0m0.219s
    user 0m0.201s
    sys 0m0.015s


    Regards,

    Rick


  4. RE: using RHEL / CentOS / Fedora perl?

    > -----Original Message-----
    > From: Rick Macdougall [mailto:rickm@ummm-beer.com]
    > Sent: Monday, September 08, 2008 10:03 AM
    > To: Justin Mason
    > Cc: users@spamassassin.apache.org
    > Subject: Re: using RHEL / CentOS / Fedora perl?
    >
    > I can verify that CentOS release 4.6 (Final) with perl, v5.8.5 built
    > for i386-linux-thread-multi is NOT affected.
    >
    > time perl perlbug.pl
    > ..................................................
    > real 0m0.219s
    > user 0m0.201s
    > sys 0m0.015s
    >
    >


    CentOS 5.2 64-bit and perl 5.10.0 rum without problems:

    # time perl perltest.pl
    .................................................. .
    real 0m0.120s
    user 0m0.096s
    sys 0m0.012s


    However, Centos 5.2 32-bit and perl 5.8.8 chokes:

    time perl perltest.pl
    .................................................. .
    real 0m4.312s
    user 0m4.272s
    sys 0m0.036s


    Jason A. Bertoch
    Network Administrator
    jason@electronet.net
    Electronet Broadband Communications
    3411 Capital Medical Blvd.
    Tallahassee, FL 32308
    (V) 850.222.0229 (F) 850.222.8771


  5. RE: using RHEL / CentOS / Fedora perl?

    Jason Bertoch wrote:
    > However, Centos 5.2 32-bit and perl 5.8.8 chokes:
    >
    > time perl perltest.pl
    > ..................................................
    > real 0m4.312s
    > user 0m4.272s
    > sys 0m0.036s


    As does CentOS 5.2 64-bit and perl 5.8.8:

    # time perl perltest.pl
    .................................................. .
    real 0m5.873s
    user 0m4.976s
    sys 0m0.021s

    Cheers,

    Phil
    --
    Phil Randal
    Networks Engineer
    Herefordshire Council
    Hereford, UK


  6. Re: using RHEL / CentOS / Fedora perl?

    Justin Mason wrote:
    > have you seen this?
    >
    > http://blog.vipul.net/2008/08/24/red...hat-a-tragedy/
    >
    > That bug in Red Hat perl will almost definitely slow down SpamAssassin,
    > too, I would say. Can anyone verify?
    >
    > --j.
    >
    >

    I don't notice any difference between my RHEL 4 (not affected) and my
    RHEL 5 (affected) installs. I am running SA through MailScanner, so I
    may not be using parts that would be affected. However, if the counts
    are small enough, the time differential is so small it would not be
    noticeable to me on our relatively busy boxes. That being said, if you
    change that 50,000 in the test loop to 500,000, it then spikes to 40
    minutes for the first half of the loop (at which point I gave up),
    instead of 4 seconds for 50,000. It does seem to exponentially slow down
    as the number of blesses called goes up.


  7. Re: using RHEL / CentOS / Fedora perl?

    Justin Mason wrote:
    > have you seen this?
    >
    > http://blog.vipul.net/2008/08/24/red...hat-a-tragedy/
    >
    > That bug in Red Hat perl will almost definitely slow down SpamAssassin,
    > too, I would say. Can anyone verify?
    >
    > --j.
    >


    This fixed it for me on a couple of centos servers:

    http://people.centos.org/z00dax/bz379791/


  8. RE: using RHEL / CentOS / Fedora perl?

    Andrew Hearn wrote:
    > Justin Mason wrote:
    >> have you seen this?
    >>
    >> http://blog.vipul.net/2008/08/24/red...hat-a-tragedy/
    >>
    >> That bug in Red Hat perl will almost definitely slow down
    >> SpamAssassin, too, I would say. Can anyone verify?
    >>
    >> --j.
    >>

    >
    > This fixed it for me on a couple of centos servers:
    >
    > http://people.centos.org/z00dax/bz379791/


    Did you notice a real-world performance boost after doing that? Got any
    numbers for pre- and post- spamassassin performance?

    Phil

    --
    Phil Randal
    Networks Engineer
    Herefordshire Council
    Hereford, UK


  9. Re: using RHEL / CentOS / Fedora perl?

    Randal, Phil wrote:
    > Andrew Hearn wrote:
    >> Justin Mason wrote:
    >>> have you seen this?
    >>>
    >>> http://blog.vipul.net/2008/08/24/red...hat-a-tragedy/
    >>>
    >>> That bug in Red Hat perl will almost definitely slow down
    >>> SpamAssassin, too, I would say. Can anyone verify?
    >>>
    >>> --j.
    >>>

    >> This fixed it for me on a couple of centos servers:
    >>
    >> http://people.centos.org/z00dax/bz379791/

    >
    > Did you notice a real-world performance boost after doing that? Got any
    > numbers for pre- and post- spamassassin performance?
    >


    No, not that I've noticed yet anyway ;-)


+ Reply to Thread