onefang said on Sun, 15 Oct 2023 08:22:52 +1000

>On 2023-10-13 09:09:51, Steve Litt wrote:
>> tito via Dng said on Thu, 12 Oct 2023 08:38:51 +0200

>> This "test" method sounds like a kludge, but it's actually better
>> because it tests what you need, not what the daemon said when it
>> supposedly became effective. If you need MariaDB running, you can
>> test with a read for a table you know is in a known database,
>> instead of hoping systemd's right about saying that it's running. So
>> in the tito file you need test scripts for each dependency, and each
>> test script is probably about 1 to 3 lines, but they are code, not
>> key value pairs.
>
>You might be forgetting about ordering? Run things in the correct
>order, then you wont have endless testing scripts all waiting for
>their "needs" to start up. Sure use the testing scripts is a good
>idea, but ALSO run things in the correct order, then the testing
>scripts wont have to wait as long.

:-) Technically you're correct. Technically, what I describe is like a
field of mousetraps with ping pong balls mounted on each, with some
traps wired so they can't snap until specific others are tested to have
snapped. It would SEEM this would take an eternity to boot, and that
indeterminate boot order would cause problems and race conditions. I
sure expected eternal boot, race conditions, and side effects from
indeterminate boot order when I started using runit in 2015.

But in fact, I never saw any problems, and boot speed is commensurate
with the other inits.

You sound like a candidate for s6. With modern additions, s6 has
determinate boot and still has all the advantages of runit (except
ultra-simplicity).

SteveT

Steve Litt

Autumn 2023 featured book: Rapid Learning for the 21st Century
http://www.troubleshooters.com/rl21