Tentative systemd slides

I recently spent some time updating my systemd knowledge and decided to put together some slides that I’ll use for a lecture. I’m interested in feedback about things that are missing,  unclear, etc. Available on slideshare, as PDF, and as LaTeX source.

4 thoughts on “Tentative systemd slides

  1. The parts about socket activation need a rewrite.

    Slide 4 description “Socket activation (on-demand startup of services)” largely misses the point; in most cases, enabling the possibility of “on-demand” is just a side benefit of socket activation.

    Slide 25 “The service is not started automatically at boot” and “systemd starts the service to handle the connection” are misleading. In the _typical_ case on most distributions, services are started automatically at boot regardless of whether they support socket activation or not.

    Using ssh as an example of socket activation on slides 26/27 is a bad choice, because it uses the non-default “Accept=yes” mode with per-connection sshd@ instances. If there’s a single example, it should be about the more normal type of service where a single service instance is started to handle the listening socket.

    One of the important features of socket activation, that of avoiding the need to specify explicit ordering for service startup, is not really mentioned except the hard-to-understand “delay synchronization” remark on slide 25. This also means that most services do not need to have any special code to tell systemd when their listening socket is up (which is usually the point at which depending services can be started).

    Socket activation also gives a way to specify socket options with a shared syntax for any service using it, without needing to reimplement the socket code for every service.

  2. @uau: I’ve kept the ssh example because it’s a way to mention instanciated services. But I’ve added an example with dovecot before it.

  3. Hi Lucas,

    Thanks for the slides – they’re very helpful.

    A few comments:

    I’m not sure “Monotonic” is the right adjective to describe the timers on page 22 – though I see the man page uses that adjective. Time is monotonic, the adjective doesn’t quite expound on what those timers are or do. Perhaps just go with “Relative timers” and leave off the “relative to different starting points”. Food for thought; you’ll make the right decision. :)

    s/instanciated/instantiated/

    Thanks again for this nice set of slides!

    -m

Comments are closed.