Detect and cache the Ping-Body-Limit response header and support non-default limit

[dimo414]

As of v2.1 responses include a Ping-Body-Limit header indicating how large the log payload can be. We could cache this value somewhere like /tmp/.healthchecks-<base_url_slug>.Ping-Body-Limit to be used in subsequent invocations.

Also, the existing hard-coded value doesn't support other installs with different limits, so a --body_limit flag should be added so callers can proactively specify the limit they'd like to use.

[bdd]

Another option would be adding start ping support to task-mon and grabbing the instance's ping body limit from that response, so you don't need to maintain any local state.

It was designed impromptu in bdd/runitor#32 (reply in thread).
Relevant lines in runitor: https://github.com/bdd/runitor/blob/main/cmd/runitor/main.go#L253-L270.

FWIW, even without instance config sniffing capability, sending start pings provide a ton of value. e.g.

• measuring the execution time
• catching task runner bugs ;)
• coupled with run ids, an ability to have multiple runners reporting to same uuid or pingKey+slug
• catching possibly stuck or unhealthily slow jobs by giving them if they don't a report status grace_time after the start ping.

[dimo414]

adding start ping support to task-mon

Start pings are already supported via the --time flag; I'd intentionally made it an opt-in feature, but I do take your point about the benefits of turning it on by default. I'll split that out into a separate FR.

coupled with run ids

I was not aware of that new feature! Very cool!

if they don't a report status grace_time after the start ping.

IIRC that was part of my motivation for not enabling --time by default. I didn't want users to need to configure the grace time properly if it didn't align with their use case.