autoheal: opt-in mode (AUTOHEAL_CONTAINER_LABEL all->autoheal)
Fleet-wide fix matching the vm-db precedent (e3a0ab5). In all-mode autoheal restarts ANY container Docker marks unhealthy, so a healthcheck that blips under load (redis-cli ping timing out) triggers an endless restart loop of a healthy container -- observed on docs 2026-08-14 (paperless-redis + docs_watchtower). Opt-in heals only containers labelled autoheal=true; real crashes stay covered by restart: unless-stopped. Co-Authored-By: Claude Opus 4.8 <[email protected]>
This commit is contained in:
co-authored by
Claude Opus 4.8
parent
d1a76f9961
commit
c08903d2be
@@ -173,5 +173,5 @@ Pulled from Infisical at deploy time (`infisical.vault.read_secrets` in `site.ym
|
|||||||
- Watchtower exposes its HTTP API on port 8088 inside the LXC (matches the
|
- Watchtower exposes its HTTP API on port 8088 inside the LXC (matches the
|
||||||
fleet pattern; 8080 collides with potential second app on this CT).
|
fleet pattern; 8080 collides with potential second app on this CT).
|
||||||
- An `autoheal` sidecar restarts unhealthy containers
|
- An `autoheal` sidecar restarts unhealthy containers
|
||||||
(`AUTOHEAL_CONTAINER_LABEL=all`) — fleet ops-sidecar standard, so the compose
|
(`AUTOHEAL_CONTAINER_LABEL=autoheal`) — fleet ops-sidecar standard, so the compose
|
||||||
is 6 containers total (app/db/redis/sidekiq + autoheal + watchtower).
|
is 6 containers total (app/db/redis/sidekiq + autoheal + watchtower).
|
||||||
|
|||||||
@@ -166,7 +166,7 @@ services:
|
|||||||
container_name: autoheal
|
container_name: autoheal
|
||||||
restart: unless-stopped
|
restart: unless-stopped
|
||||||
environment:
|
environment:
|
||||||
- AUTOHEAL_CONTAINER_LABEL=all
|
- AUTOHEAL_CONTAINER_LABEL=autoheal
|
||||||
- AUTOHEAL_INTERVAL=15
|
- AUTOHEAL_INTERVAL=15
|
||||||
- AUTOHEAL_START_PERIOD=60
|
- AUTOHEAL_START_PERIOD=60
|
||||||
- AUTOHEAL_DEFAULT_STOP_TIMEOUT=30
|
- AUTOHEAL_DEFAULT_STOP_TIMEOUT=30
|
||||||
|
|||||||
Reference in New Issue
Block a user