haproxy process check should not go hard critical before hard-stop-after

[Paul Collins]

It looks like the haproxy process check can go hard critical before hard-stop-after has had a chance to do its work.

Our defaults are as follows (these numbers are minutes):

        max_check_attempts 3
        normal_check_interval 5
        retry_check_interval 1

So we check OK services every 5 minutes, on the first CRITICAL we start checking every 1 minute until it's reaches three failures, which sends it HARD CRITICAL after 3 minutes. But we set hard-stop-after to 5m.

The values above can be defined in any service config file, so the charm could override them to allow the extra N minutes for hard-stop-after to take effect.

Probably the simplest thing to do is to have the check config do the following:

        max_check_attempts N+2

N+1 might work, but I suspect that relies on everything working too well, and an extra minute isn't the end of the world.