amayer1983/docksentry v1.17.2

v1.17.2 — Smarter healthcheck for slow-startup apps

on GitHub

Fixed

Slow-startup apps (GitLab, Nextcloud, Mastodon, …) auto-rolled-back every cron tick

The health-check timeout was 5 minutes, and a container still in state=running, health=starting after that was treated as failure → automatic rollback. GitLab routinely needs 8-15 minutes for first-boot migrations and Rails warm-up. Result: the user got a "Health check failed — rolled back" notification every evening and GitLab never actually updated.

Three changes to the wait logic:

1. Default wait raised 300s → 600s (HEALTHCHECK_MAX_STARTING env var, configurable).
2. Respect the image's own Healthcheck.StartPeriod — the effective wait is max(default, start_period × 1.5). An image declaring start_period: 5m no longer gets cut off after our default.
3. state=running AND health=starting after the wait → no rollback. Container is alive but slow; Docker's own healthcheck keeps running and will eventually flip the bit. Reported as a soft success (⚠ updated but still 'starting' — left running) so the rest of the batch continues.

Active unhealthy or not-running still rolls back as before (standalone path) — only the over-eager "still starting → failure" verdict is gone.

Compose-path "rolled back" messages were misleading

For compose containers the previous "rollback" was a no-op (the same compose file produces the same container). The message now says container left in place (compose) honestly instead of claiming a rollback that didn't happen.

Added

Last 10 log lines attached to health-check notifications

When a container is rolled back or reports as "still starting", the Telegram / Discord / webhook message now includes a code-fenced tail of its logs — so you can see in chat what was happening without SSH-ing to the host:

⚠ gitlab: Updated but still 'starting' after our wait — left running.
Last logs:
2026-05-27 19:01:23 ==> /var/log/gitlab/rails/application.log <==
2026-05-27 19:01:25 Running database migrations...

HEALTHCHECK_MAX_STARTING env var (default 600s)

For users with super-slow apps that need an even longer ceiling. Persisted via settings.json (advanced setting).

docker pull amayer1983/docksentry:latest
docker compose up -d