Detect node start using healthz

Use the operations service health check to detect when nodes are started
in test-network-nano-bash instead of using arbitrary sleeps. Depending
on the host machine and whether CAs are used, this can reduce network
start time by up to 15 seconds.

Also ensure the Docker socket location is set correctly for Docker
contexts that use a non-default location. This is required to prevent
the health service from returning Docker connection errors.

Signed-off-by: Mark S. Lewis <Mark.S.Lewis@outlook.com>

test-network-nano-bash/network.sh

@@ -14,7 +14,6 @@ printHelp() {
     echo "  Starts the test network"
     echo
     echo "    Flags:"
-    echo "    -d <delay>         - CLI delays for a certain number of seconds (defaults to 3)"
     echo "    -o <orderer_type>  - Specify the orderer type. BFT or etcdraft. (defaults to etcdraft)"
     echo "    -ca                - Use CAs instead of cryptogen.  (defaults to cryptogen)"
     echo "    -h                 - Print this message"
@@ -49,8 +48,6 @@ networkStop() {
 }
 
 networkStart() {
-  : "${CLI_DELAY:=5}"
-
   # shellcheck disable=SC2064
   trap networkStop 0 1 2 3 15
 
@@ -62,8 +59,9 @@ networkStart() {
     ./ordererca.sh  > ./logs/ordererca.log 2>&1 &
     ./org1ca.sh  > ./logs/org1ca.log 2>&1 &
     ./org2ca.sh  > ./logs/org2ca.log 2>&1 &
-    echo "Waiting ${CLI_DELAY}s..."
-    sleep "${CLI_DELAY}"
+
+    echo 'Waiting for CAs to start...'
+    waitForHealthzOK 9843 9844 9845
   fi
 
   if [ -d "${PWD}"/channel-artifacts ] && [ -d "${PWD}"/crypto-config ]; then
@@ -84,13 +82,17 @@ networkStart() {
   ./orderer2.sh "${ORDERER_TYPE}" > ./logs/orderer2.log 2>&1 &
   ./orderer3.sh "${ORDERER_TYPE}" > ./logs/orderer3.log 2>&1 &
 
+  ORDERER_HEALTHZ_PORTS="8443 8444 8445"
+
   #start one additional orderer for BFT consensus
   if [ "$ORDERER_TYPE" = "BFT" ]; then
     ./orderer4.sh "${ORDERER_TYPE}" > ./logs/orderer4.log 2>&1 &
+    ORDERER_HEALTHZ_PORTS="${ORDERER_HEALTHZ_PORTS} 8450"
   fi
 
-  echo "Waiting ${CLI_DELAY}s..."
-  sleep "${CLI_DELAY}"
+  echo 'Waiting for orderers to start...'
+  # shellcheck disable=SC2086
+  waitForHealthzOK ${ORDERER_HEALTHZ_PORTS}
 
   echo "Starting peers..."
   ./peer1.sh > ./logs/peer1.log 2>&1 &
@@ -98,8 +100,8 @@ networkStart() {
   ./peer3.sh > ./logs/peer3.log 2>&1 &
   ./peer4.sh > ./logs/peer4.log 2>&1 &
 
-  echo "Waiting ${CLI_DELAY}s..."
-  sleep "${CLI_DELAY}"
+  echo 'Waiting for peers to start...'
+  waitForHealthzOK 8446 8447 8448 8449
 
   if [ "${CREATE_CHANNEL}" = "true" ]; then
     echo "Joining orderers to channel..."
@@ -126,6 +128,25 @@ networkStart() {
   wait
 }
 
+waitForHealthzOK() {
+  while [ $# -ge 1 ]; do
+    HEALTHZ_STATUS=
+    HEALTHZ_PORT="$1"
+    shift
+
+    while [ "${HEALTHZ_STATUS}" != 'OK' ]; do
+      # Sleep before a retry
+      if [ "${HEALTHZ_STATUS}" ]; then
+        [ "${HEALTHZ_STATUS}" != 'OFFLINE' ] && echo "Healthz port ${HEALTHZ_PORT}: ${HEALTHZ_RESPONSE}"
+        sleep 0.5
+      fi
+
+      HEALTHZ_RESPONSE=$(curl --silent "http://127.0.0.1:${HEALTHZ_PORT}/healthz" || echo '{"status":"OFFLINE"}')
+      HEALTHZ_STATUS=$(echo "${HEALTHZ_RESPONSE}" | jq --raw-output '.status')
+    done
+  done
+}
+
 networkClean() {
   echo "Removing directories: channel-artifacts crypto-config data logs"
   rm -r "${PWD}"/channel-artifacts || true
@@ -142,7 +163,7 @@ if [ $# -lt 1 ] ; then
   printHelp
   exit 0
 else
-  MODE=$1
+  MODE="$1"
   shift
 fi
 
@@ -153,13 +174,19 @@ INCLUDE_CA=false
 while [ $# -ge 1 ] ; do
   key="$1"
   case $key in
-  -d )
-    CLI_DELAY="$2"
-    shift
-    ;;
   -o )
-    ORDERER_TYPE="$2"
+    ORDERER_TYPE="$(echo "$2" | tr '[:upper:]' '[:lower:]')"
     shift
+
+    if [ "${ORDERER_TYPE}" = bft ]; then
+      ORDERER_TYPE=BFT
+    fi
+
+    if [ "${ORDERER_TYPE}" != etcdraft ] && [ "${ORDERER_TYPE}" != BFT ]; then
+      echo "Unsupported orderer type: ${ORDERER_TYPE}" >&2
+      printHelp "$MODE"
+      exit 1
+    fi
     ;;
   -ca )
     INCLUDE_CA=true
@@ -177,6 +204,15 @@ while [ $# -ge 1 ] ; do
   shift
 done  
 
+# Ensure peers are configured with the correct docker socket location, otherwise
+# health checks will fail.
+export CORE_VM_ENDPOINT
+if [ -z "${DOCKER_HOST:-}" ]; then
+  CORE_VM_ENDPOINT="$(docker context inspect --format '{{.Endpoints.docker.Host}}')"
+else
+  CORE_VM_ENDPOINT="${DOCKER_HOST}"
+fi
+
 if [ "$MODE" = "start" ]; then
   networkStart
 elif [ "$MODE" = "clean" ]; then