From e215725a6170731a9508f92e654794a1892b8fba Mon Sep 17 00:00:00 2001 From: Michael Prokop Date: Sat, 6 May 2017 01:02:25 +0200 Subject: [PATCH] FAQ: provide information about switch to systemd --- faq/index.html.tt2 | 57 +++++++++++++++++++++++++++++++++++++++++++++++++++++- 1 file changed, 56 insertions(+), 1 deletion(-) diff --git a/faq/index.html.tt2 b/faq/index.html.tt2 index 01b3ba0..3fbe85e 100644 --- a/faq/index.html.tt2 +++ b/faq/index.html.tt2 @@ -28,7 +28,7 @@

FAQ for grml

-

Up2date: applies to Grml version 2014.11

+

Up2date: applies to Grml version 2017.05

Index:

@@ -42,6 +42,7 @@
  • What about the release name?
  • Requirements for running Grml
  • Which bootoptions does Grml support?
  • +
  • Why is Grml using systemd?
  • Are there any known issues with this release? How about reporting bugs?
  • @@ -151,6 +152,60 @@ href="http://www.kernel.org/doc/Documentation/kernel-parameters.txt">kernel-parameters.txt of the Linux kernel applies to Grml as well.

    +

    Why is Grml using systemd?

    + +

    The switch from file-rc to systemd happened for various reasons. + Grml used file-rc for many years, mainly because it provided a better + way to control startup behavior via its /etc/runlevel.conf configuration + than with using sysvinit. Though for us Grml developers this also meant + that whenever there have been any changes in Debian's startup + configuration we had to compare our /etc/runlevel.conf setup with what a + normal Debian system would give us. Users who wanted to remaster Grml + with a custom startup procedure as well had to practically fork + maintenance of the /etc/runlevel.conf file. This didn't only mean + tracking new features/services, but also solve any possible issues + around it - duplicating efforts and wasting developers time + unnecessarily. Lately we also started to see problems that no one else + seemed to have (or care about enough), for example with multiple network + cards we ran into race-conditions with resolvconf. Problems like that + turned out to be release stoppers for us.

    + +

    systemd on the other hand provides great documentation, service + supervision, takes care of parallel service startup and is the default + init system on most Linux distributions nowadays. This means more users, + better testing and integration. Logging, startup time investigation (to + get a fast boot procedure) and identifying failed service startups with + sysvinit/file-rc was always hard, unreliable or even impossible under + certain conditions. bootlogd was unreliable (while `journalctl -b` is + available out-of-the-box with systemd), bootchart not nicely integrated + (while systemd-analyze blame/critical-chain works out-of-the-box) and we + aren't aware of any equivalence for e.g. `systemctl --failed`.

    + +

    While migrating our stack to systemd is not fully finished + yet, its switch - at least so far - was easier than expected. It also + turns out that it gives users who want to remaster Grml (or build their + very own ISOs from scratch using grml-live) more flexibility and control + over the startup process. systemd's override.conf mechanism and preset + feature provides the flexibility to overwrite unwanted behavior, without + losing the option to use existing defaults.

    + +

    Last but not least we think it's good that systemd is actively + maintained and receives attention. The sysvinit/file-rc ecosystem was + stagnating/non-existend for too many years. Grml used its own initrd + implementation in its very beginnings, until a more broadly available + initramfs-tools/ live-boot solution appeared, broadening the user base, + sharing goals amongst different (live) distributions. Back in the days + Grml - like many other live distributions - had to implement hardware + recognition on its own. While udev received lots of complaints back + then, its integration actually solved all the hardware recognition for + the good. systemd's vision of stateless systems is something which helps + building live systems like Grml.

    + +

    While we don't claim that systemd is perfect and doesn't have its + issues and drawbacks (like any software), we're happy about its + existence and more than happy about development and support by Debian's + systemd folks.

    +

    Are there any known issues? How about reporting bugs?

    -- 2.1.4