On Thu, 22 Jun 2023, Mason Loring Bliss wrote:
On Fri, Jun 23, 2023 at 01:30:30AM +0200, Thorsten Glaser wrote:
On Thu, 22 Jun 2023, David Griffith wrote:
This was prompted when I found that rsyslog stopped working on Bullseye when
upgraded to Bookworm. What sort of depenency would you suggest to implement
the following?
“Read the release notes.”
What part of the release notes are you referring to? It might be useful to
reference that specifically, in addition to the pointer.
Nothing relevant is obvious here for rsyslogd:
https://www.debian.org/releases/bookworm/amd64/release-notes/index.en.html
We probably want to avoid nasty surprises. I haven't upgraded my servers to
Bookworm yet, but scanning the release notes, nothing there would indicate
to me that my logging services were about to break.
Exactly. This problem started off as a violation of the basic rule of
least surprise. When I upgrade an operating system, I don't want to see
fundamental services not come back. Stuff like that should be fixed
before release. And if found after release, a fix should not be delayed
until the next major release.
David's suggestion of requiring orphan-sysvinit-scripts seems like a
sensible solution that will prevent outages. Let's posit, for instance,
that there's a hospital somewhere using Debian for some of their
infrastructure, and running sysvinit because they require reliable systems.
It would seem responsible to not arbitrarily let logging break on an
upgrade. That's not what people expect from Debian.
While a life-and-death system shouldn't be running a standard OS at all,
but instead something realtime, that shouldn't cause anyone to pooh-pooh
this issue. No otherwise reliable operating system (RT or not) should
have things silently stop working without alerting the admin/user of said
things.
--
David Griffith
d...@661.org
