Why does it take about "many" 5-minute-cycles to get a "two-minute DOWN service" as UP again?

[thomasmerz]

I have a daily update-policy for my docker-based services: Pulling, then checking if newer image has been downloaded and if this is the case: stop and start docker container. Usually this takes about less than two minute(s) (also in this case below) until "ready":

https://thomasmerz.github.io/upptime/incident/29

$ git log --grep Pihole-HEL

…
commit b873a31e1c7664c4a48eaf5d1724d82b62ca3cbf
Author: Upptime Bot <73812536+upptime-bot@users.noreply.github.com>
Date:   Mon Feb 1 06:57:04 2021 +0000

    🟩 Pihole-HEL is up (200 in 1409 ms) [skip ci] [upptime]

commit fa4d4fe5dcd28e621680864d3b9d450a2164473e
Author: Upptime Bot <73812536+upptime-bot@users.noreply.github.com>
Date:   Mon Feb 1 06:30:18 2021 +0000

    🟥 Pihole-HEL is down (0 in 0 ms) [skip ci] [upptime]
…

First I thought, this might be due to my .upptimerc.yaml

workflowSchedule:
  graphs: "*/5 * * * *"
  responseTime: "*/5 * * * *"
  staticSite: "*/5 * * * *"
  summary: "*/5 * * * *"
  updateTemplate: "0 0 * * *"
  updates: "0 3 * * *"
  uptime: "*/5 * * * *"

But everything "important" should run every 5 minutes… 🤔

Q: Why does upptime need so much 5-minute cycles to determine the service as UP again? I expected that it would has been UP within the next 5-minute cycle… (which has been indeed started again at 07:31:52 after being stopped at 07:30:11)

[AnandChowdhary]

This is because the next Upptime workflow ran after 25 minutes, not after 5 minutes. It actually did indeed mark your site as up the next time Uptime CI ran, but the problem is that GitHub sometimes runs your CI slower than every 5 minutes. Sometimes, it could be as slow as once every hour, sometimes it runs every 5 minutes.

You can see here that indeed Uptime CI is running slower: https://github.com/thomasmerz/upptime/actions?query=workflow%3A%22Uptime+CI%22

For more details: https://upptime.js.org/blog/2021/01/22/github-actions-schedule-not-working

[thomasmerz]

Quick solution: I will turn down my local check-freqeuency (5 -> 15 min), to minimize so long persisting false positive DOWNs.
Long-term solution: I might need to get this REST API up and running for my workflow…