[systemd-devel] Support for large applications

[Avi Kivity]

On 02/19/2016 03:24 PM, Michal Sekletar wrote:
> On Fri, Feb 19, 2016 at 1:49 PM, Zbigniew Jędrzejewski-Szmek
> <zbyszek at in.waw.pl> wrote:
>
>> I don't think there's a way around the issue short of allowing
>> watchdog during startup. Databases which do long recovery are a bit
>> special, most programs don't exhibit this kind of behaviour, but maybe
>> this case is important enough to add support for it.
> Rather than configuring huge start timeout for the service, firing it
> up and hoping for the best I was thinking about disabling start
> timeout entirely, make service of type "notify" and signal service
> readiness via sd_notify. Also update progress by sending STATUS
> message when some significant stage of DB startup procedure finishes.
> This is not solving watchdog problem, but I think it would be an
> improvement anyway. At least, it would provide better debugability of
> start-up of this particular service.
>

We do use sd_notify (and report incremental progress), but we what if 
startup hangs?  Recall that our startup procedure may be very long.

We want, on one hand, systemd to detect a hung process, and on the other 
hand, allow for long log replays.  The only solution I can see is a 
watchdog that monitors startup.