Supervise ckpool with restart loop and bitcoind readiness check

Replaces the one-shot ckpool launch with a supervised loop that waits
for bitcoind to be reachable before each start. When the API kills
ckpool (bitcoind down), the loop waits for recovery and restarts it
automatically — dashboard stays up the whole time.

docker_entrypoint.sh

@@ -136,10 +136,6 @@ sed \
     -e "s|\${ZMQ_BLOCK}|${ZMQ_BLOCK}|g" \
     "${TEMPLATE}" > "${CONF}"
 
-echo "kamado-entrypoint: starting ckpool (solo, ${BITCOIND_VARIANT}) on port ${STRATUM_PORT}, loglevel ${CKPOOL_LOGLEVEL}"
-/usr/local/bin/ckpool --btcsolo --config "${CONF}" --sockdir "${SOCKET_DIR}" --log-shares -l "${CKPOOL_LOGLEVEL}" &
-CKPOOL_PID=$!
-
 # DB_PATH must live on the persisted main volume; the default
 # /var/lib/kamado/kamado.db is ephemeral container storage.
 KAMADO_DATA_DIR=/root/.kamado/data
@@ -161,6 +157,42 @@ echo "kamado-entrypoint: starting kamado-api"
 /usr/local/bin/kamado-api &
 API_PID=$!
 
+# wait_for_bitcoind blocks until bitcoind responds to getblockchaininfo.
+# Called before each ckpool start so we don't launch ckpool into a wall.
+wait_for_bitcoind() {
+    local url="http://${BITCOIN_RPC_USER}:${BITCOIN_RPC_PASSWORD}@${BITCOIN_RPC_HOST}:${BITCOIN_RPC_PORT}"
+    local backoff=2
+    while true; do
+        if curl -sf --max-time 5 \
+            -d '{"jsonrpc":"1.0","method":"getblockchaininfo","params":[]}' \
+            -H 'Content-Type: application/json' \
+            "${url}" >/dev/null 2>&1; then
+            return 0
+        fi
+        echo "kamado-entrypoint: waiting for bitcoind (retry in ${backoff}s)..."
+        sleep "${backoff}"
+        backoff=$(( backoff < 30 ? backoff * 2 : 30 ))
+    done
+}
+
+# Supervised ckpool restart loop. When ckpool exits (killed by the API
+# on bitcoind failure, or crashed), we wait for bitcoind to be reachable
+# again before restarting. This keeps ckpool alive when bitcoind is
+# healthy and lets miners failover when it's not — without restart-
+# looping the entire container.
+run_ckpool_loop() {
+    while true; do
+        wait_for_bitcoind
+        echo "kamado-entrypoint: starting ckpool (solo, ${BITCOIND_VARIANT}) on port ${STRATUM_PORT}, loglevel ${CKPOOL_LOGLEVEL}"
+        /usr/local/bin/ckpool --btcsolo --config "${CONF}" --sockdir "${SOCKET_DIR}" --log-shares -l "${CKPOOL_LOGLEVEL}"
+        EXIT_CODE=$?
+        echo "kamado-entrypoint: ckpool exited (code ${EXIT_CODE}), will restart after bitcoind is reachable"
+        sleep 2
+    done
+}
+run_ckpool_loop &
+CKPOOL_LOOP_PID=$!
+
 # Optional TLS stratum via stunnel sidecar.
 STUNNEL_PID=""
 if [[ "${TLS_ENABLED}" == "enabled" ]]; then
@@ -305,16 +337,20 @@ fi
 
 term() {
     echo "kamado-entrypoint: SIGTERM — shutting down"
-    kill -TERM "${API_PID}" "${CKPOOL_PID}" ${STUNNEL_PID:-} 2>/dev/null || true
-    wait "${API_PID}" "${CKPOOL_PID}" ${STUNNEL_PID:-} 2>/dev/null || true
+    kill -TERM "${API_PID}" "${CKPOOL_LOOP_PID}" ${STUNNEL_PID:-} 2>/dev/null || true
+    # Kill any running ckpool process inside the loop.
+    pkill -TERM -f '/usr/local/bin/ckpool' 2>/dev/null || true
+    wait "${API_PID}" "${CKPOOL_LOOP_PID}" ${STUNNEL_PID:-} 2>/dev/null || true
     exit 0
 }
 trap term TERM INT
 
-# shellcheck disable=SC2086
-wait -n ${CKPOOL_PID} ${API_PID} ${STUNNEL_PID:-}
+# The API is the critical process — if it exits, the container should
+# restart. The ckpool loop manages its own lifecycle independently.
+wait "${API_PID}"
 EXIT_CODE=$?
-echo "kamado-entrypoint: a supervised process exited (${EXIT_CODE}), stopping the rest"
-kill -TERM "${API_PID}" "${CKPOOL_PID}" ${STUNNEL_PID:-} 2>/dev/null || true
+echo "kamado-entrypoint: kamado-api exited (${EXIT_CODE}), stopping the rest"
+kill -TERM "${CKPOOL_LOOP_PID}" ${STUNNEL_PID:-} 2>/dev/null || true
+pkill -TERM -f '/usr/local/bin/ckpool' 2>/dev/null || true
 wait || true
 exit "${EXIT_CODE}"