(podman) rollback systemd unit watchdog feature

Author: zeylos

roles/podman/files/application_execstart.sh

@@ -5,8 +5,6 @@ set -euo pipefail
 TIMEOUT="${TIMEOUT:-300}"
 INTERVAL="${INTERVAL:-5}"
 
-# --- Phase 1: Startup health wait ---
-
 # Read container IDs into array, one per line, then put it into the CONTAINERS var
 mapfile -t CONTAINERS < <(podman-compose ps -q)
 TOTAL=${#CONTAINERS[@]}
@@ -46,32 +44,3 @@ while true; do
 done
 
 systemd-notify --ready --status="All ${TOTAL} containers healthy"
-
-# --- Phase 2: Watchdog loop ---
-
-# WATCHDOG_USEC is set automatically by systemd from WatchdogSec= (in microseconds).
-# We convert to seconds and halve it so we actually check in well before the deadline.
-# We also hardcode on purpose the minimal healcheck interval to 5 seconds.
-WATCHDOG_USEC="${WATCHDOG_USEC:-60000000}"
-WATCHDOG_SEC=$((WATCHDOG_USEC / 1000000 / 2))
-[ "$WATCHDOG_SEC" -lt 5 ] && WATCHDOG_SEC=5
-
-while true; do
-  sleep "$WATCHDOG_SEC"
-
-  for c in "${CONTAINERS[@]}"; do
-    STATE=$(podman inspect --format "{{.State.Status}}" "$c" 2>/dev/null || true)
-    if [ "$STATE" != "running" ]; then
-      echo "Container $c is no longer running (state: ${STATE:-gone})"
-      exit 1
-    fi
-
-    HEALTH=$(podman inspect --format "{{.State.Health.Status}}" "$c" 2>/dev/null || true)
-    if [ "$HEALTH" = "unhealthy" ]; then
-      echo "Container $c became unhealthy"
-      exit 1
-    fi
-  done
-
-  systemd-notify WATCHDOG=1 --status="Watchdog: ${TOTAL} containers healthy"
-done

roles/podman/templates/application_systemd_unit.j2

@@ -24,9 +24,5 @@ Environment=TIMEOUT={{ st_podman_application_timeout }}
 # TIMEOUT + 30s buffer for podman-compose overhead (pulling, network setup, etc.)
 TimeoutStartSec={{ st_podman_application_timeout + 30 }}
 
-# How often systemd expects a heartbeat from the ExecStart script.
-# If no check-in within this window, systemd considers the service hung.
-WatchdogSec=60
-
 [Install]
 WantedBy=default.target