[Date Prev][Date Next] [Chronological] [Thread] [Top]

RE: idletimeout implementation



> -----Original Message-----
> From: owner-openldap-devel@OpenLDAP.org
> [mailto:owner-openldap-devel@OpenLDAP.org]On Behalf Of
> HORSTMAN, MARK A (SBCSI)

> I can't figure out why global_idletimeout is divided by 4
> before it's added
> to the last_idle_check value (/SLAPD_IDLE_CHECK_LIMIT) in
> daemon.c.  Doesn't
> this cause the idle timeout to actually be 1/4th the configured value?

No, but it causes idle times to be checked 4 times more often. Since
connections can be established at any time, if idle checks were only
performed once per idle_check value, you may get a connection that actually
hangs around for idle_check*2 seconds before it's detected. By checking 4
times more often, the slop is reduced to idle_check*1.25. This is just basic
sampling/oversampling theory at work, nothing special.

  -- Howard Chu
  Chief Architect, Symas Corp.       Director, Highland Sun
  http://www.symas.com               http://highlandsun.com/hyc
  Symas: Premier OpenSource Development and Support