[plug] Systemd, good or bad
Brad Campbell
brad at fnarfbargle.com
Mon Sep 22 08:46:39 UTC 2014
On 22/09/14 16:28, Hani Jabr wrote:
> Slow starting apps? Why wait for a DB to initialise before starting sshd, or vice versa? I have one box that starts multiple instances of DB2 that could just as easily be started in parallel (the bottleneck is the LUNs, not the HBA).
I can see that. My boxes bring up ssh and other essentials before any
application daemons.
> When you reboot a production machine, do you really never go back and check if it is working?
>
_Really_. I have far better things to do with my time and I have a
system that will let me know if something fails to come up. No need to
waste time drumming my fingers, when I could be doing other stuff.
Let me reverse the question, do you really have systems that fragile you
need to check they are working after a reboot?
In the rare event something fails to come up (and I mean maybe once a
year in a bad year) it gets diagnosed and properly fixed. I don't have
time to paper over problems with reboots and babysitting.
I was just reading part of a blog post talking about systemd's
"advantage" of being able to automatically restart crashed daemons.
I'm chasing a memory leak in a CCTV system at the moment, and the
software has the interesting feature that if the service is restarted
more than 3 times in a 24 hour period it forces a server reboot. I asked
the developers what that horrible kludge was doing in their otherwise
well engineered application and they replied "We only run on windows,
what do you expect?".
If something crashes it's broken and needs to be fixed, not restarted.
Brad
--
Dolphins are so intelligent that within a few weeks they can
train Americans to stand at the edge of the pool and throw them
fish.
More information about the plug
mailing list