[SSSD] what is sbus_timeout?

Simo Sorce ssorce at redhat.com
Wed Nov 4 13:37:26 UTC 2009


On Wed, 2009-11-04 at 07:51 -0500, Stephen Gallagher wrote:
> -----BEGIN PGP SIGNED MESSAGE-----
> Hash: SHA1
> 
> On 11/04/2009 06:05 AM, Jakub Hrozek wrote:
> > On 11/04/2009 07:09 AM, David O'Brien wrote:
> >> David O'Brien wrote:
> >>> This is in /etc/sssd/sssd.conf in the [sssd] section. I guess it's 
> >>> something like how long should a client or connection continue to query 
> >>> a back end before timing out, but I want to make sure.
> >>>
> >>> thanks
> >> found it
> > 
> >> Specifies the timeout for messages sent over the SBUS. I don't know what 
> >> these might be; anyone care to drop an example for my own education? 
> > 
> > SBUS is an implementation of the DBUS protocol that SSSD uses internally
> > for communication between various parts of SSSD. It does not require a
> > routing daemon (a bus).
> > 
> > I'm actually thinking that if we keep these low level knobs documented
> > in man pages, there should be a short explanation of acronyms used (like
> > SBUS) either in the man page or the HOWTO.
> > 
> >> Does this have to be type "string" to allow for a default value of -1? I 
> > 
> > Man page bug, it is really of type "int"
> > 
> 
> We really need to rename this option to something more comprehensible.
> It is only used for two actions in the SSSD code:
> 
> 1) When sending a heartbeat ping to the monitor's child processes, if
> the child does not respond with it's pong within the sbus_timeout, it is
> considered to be hung.
> 
> 2) When the monitor wants to order a child to shutdown (when the SSSD is
> shutting down, or when config changes are pushed live (future feature),
> this timeout is used to determine whether the child has received the
> shutdown command.
> 
> So perhaps we should change the name to something like
> heartbeat_timeout and set the default to an explicit value.
> 
> Opinions?
> 
> 
> A bit of history: Originally, we had intended this to be the default
> timeout for all inter-process communication in the SSSD, so that's why
> it was named sbus_timeout. As development progressed, we started
> deciding that nearly every separate call needed its own timeout value,
> and this setting became increasingly less usefully defined.

I'd prefer to keep it undocumented for now, and not see it in the
configuration files.
Most people wouldn't know how to set it properly anyways.

Simo.

-- 
Simo Sorce * Red Hat, Inc * New York




More information about the sssd-devel mailing list