Loading ...
Sorry, an error occurred while loading the content.

Re: Proxymap tunning

Expand Messages
  • Victor Duchovni
    ... This naive view is inaccurate. ... You are misdirecting your efforts, and probably doing more harm than good. The operating system does a good job of
    Message 1 of 5 , Jul 1, 2010
    • 0 Attachment
      On Thu, Jul 01, 2010 at 12:39:22PM +0200, Luciana Moreira wrote:

      >> The cleanup and SMTP servers can work in parallel. If mysql is slow,
      >> make sure your tables are properly indexed, and queries are not so
      >> complex that they can only be resolved via a "table scan".
      >>
      > Is this also true for one single message?
      >
      > I know the cleanup server can process messages while smtpd is executing
      > other tasks. But I expected the flow of a single message to be smtpd and
      > then cleanup, i.e. for the same message we would not have parallel requests
      > to cleanup and sender checks.

      This naive view is inaccurate.

      > btw, mysql is not slow, it is just that we would like to avoid wasted
      > memory allocation for processes that will be _always_ sitting there doing
      > nothing since there are more processes assigned than actually needed.

      You are misdirecting your efforts, and probably doing more harm than good.
      The operating system does a good job of managing idle memory pages, and
      the process count is small enough that you'd need extremely old hardware
      for this to be an issue.

      > Another question also in regard to proxymap tuning: If I define less
      > proxymap processes than what may be needed in a certain time period, will
      > the postfix scheduler hold the request until a process is available or will
      > there be an error returned to the function using mysql?

      Don't tune proxymap, it does not need tuning.

      --
      Viktor.
    • Luciana Moreira
      Thx a lot for the reply Victor, it clarified some of my doubts. From your response to my last question I am assuming the scheduler handles the proxymap process
      Message 2 of 5 , Jul 1, 2010
      • 0 Attachment
        Thx a lot for the reply Victor, it clarified some of my doubts.

        From your response to my last question I am assuming the scheduler
        handles the proxymap process allocation and I don't need to care about
        possible errors.

        Just to clarify, you recommend to "not tune proxymap". Our system
        requirement is to always have at least 1 proxymap process ready for use
        at any given time. This means:

        1) That I would leave the default value there (100).

        2) 3 functions may need proxymap in parallel for each message

        3) Postfix would be able to handle ~33 messages in parallel (respecting
        our requirement)

        4) We would define 33 smtpd processes

        5) Defining more smtpd processes will not result in an error but will
        not respect our system requirement. Defining less smtpd processes is
        definitely a "waste" of proxymaps (even though OS can handle it).

        Did I understand your valuable input right? :)

        Why do you believe I would be possible causing harm to the system, is
        there some side effect to the changes I am making which I am unaware of?

        Regards,
        Luciana

        Victor Duchovni wrote:
        > On Thu, Jul 01, 2010 at 12:39:22PM +0200, Luciana Moreira wrote:
        >
        >
        >>> The cleanup and SMTP servers can work in parallel. If mysql is slow,
        >>> make sure your tables are properly indexed, and queries are not so
        >>> complex that they can only be resolved via a "table scan".
        >>>
        >>>
        >> Is this also true for one single message?
        >>
        >> I know the cleanup server can process messages while smtpd is executing
        >> other tasks. But I expected the flow of a single message to be smtpd and
        >> then cleanup, i.e. for the same message we would not have parallel requests
        >> to cleanup and sender checks.
        >>
        >
        > This naive view is inaccurate.
        >
        >
        >> btw, mysql is not slow, it is just that we would like to avoid wasted
        >> memory allocation for processes that will be _always_ sitting there doing
        >> nothing since there are more processes assigned than actually needed.
        >>
        >
        > You are misdirecting your efforts, and probably doing more harm than good.
        > The operating system does a good job of managing idle memory pages, and
        > the process count is small enough that you'd need extremely old hardware
        > for this to be an issue.
        >
        >
        >> Another question also in regard to proxymap tuning: If I define less
        >> proxymap processes than what may be needed in a certain time period, will
        >> the postfix scheduler hold the request until a process is available or will
        >> there be an error returned to the function using mysql?
        >>
        >
        > Don't tune proxymap, it does not need tuning.
        >
        >
      Your message has been successfully submitted and would be delivered to recipients shortly.