URL:
https://github.com/SSSD/sssd/pull/94
Title: #94: Enable {socket,dbus}-activation for responders
fidencio commented:
"""
On Tue, Nov 29, 2016 at 10:05 AM, Fabiano Fidêncio <fidencio(a)redhat.com>
wrote:
On Tue, Nov 29, 2016 at 9:33 AM, Pavel Březina <notifications(a)github.com>
wrote:
> One thing I don't like about those patches is that we always recreate the
> idle timer. Can we get around this (maybe also for the client idle
> timeout)? What I have i mind is this:
>
> 1. Instead of resetting the timer, remember time(NULL) of the last
> communication, say last_request_time
> 2. Create a periodic timer that fires each responder_idle_timeout / 2
> 3. In the timer check that last_request_time + responder_idle_timeout
> < time(NULL)
>
> It won't be as precise as your solution, but I don't think we need it to
> be but it will save us lots of memory operations. What do you think?
>
Fine by me. Your suggestion has been used in a few other places of the our
codebase as well.
After doing some tests here I do believe it may be a good approach as long
as the periodic timeout fires each 60s (that's the minimum amount of time
I'm setting up for the timeout. That looks like a good balance between
precision and number of operations, what do you think?
About the client timeout, I will do the same as it's also reset in every
communication and I also would like to use the minimum amount of time
allowed for firing for the periodic timer.
Best Regards,
--
Fabiano Fidêncio
"""
See the full comment at
https://github.com/SSSD/sssd/pull/94#issuecomment-263555469