Merge pull request #261 from EventStore/yoeight/db-622-fix-cluster-di…

…scovery-process-in-the-java-client

Fix cluster discovery process.

db-client-java/src/main/java/com/eventstore/dbclient/ConnectionService.java

@@ -142,6 +142,7 @@ public void createChannel(UUID previousId, InetSocketAddress candidate) {
 
             if (this.loadServerFeatures()) {
                 this.channelId = UUID.randomUUID();
+                this.connection.confirmChannel();
                 logger.info("Connection to endpoint [{}] created successfully", this.connection.getLastConnectedEndpoint());
                 break;
             }

db-client-java/src/main/java/com/eventstore/dbclient/ConnectionState.java

@@ -22,6 +22,12 @@ class ConnectionState {
     private InetSocketAddress previous;
     private ManagedChannel currentChannel;
 
+    // Indicates if the current channel passed all the connection pre-requisites to be used by the user
+    // Not exhaustive list includes:
+    // * If we managed to get a gossip seed from the channel
+    // * If we managed to read the server features (if not, it was a not found error then it's not fatal, just old node version)
+    private boolean confirmedChannel;
+
     ConnectionState(EventStoreDBClientSettings settings) {
         this.settings = settings;
 
@@ -46,7 +52,11 @@ class ConnectionState {
     }
 
     InetSocketAddress getLastConnectedEndpoint() {
-        return this.previous;
+        return this.confirmedChannel ? this.previous : null;
+    }
+
+    void confirmChannel() {
+        this.confirmedChannel = true;
     }
 
     ManagedChannel getCurrentChannel() {
@@ -82,6 +92,7 @@ void connect(InetSocketAddress addr) {
             builder.keepAliveTime(settings.getKeepAliveInterval(), TimeUnit.MILLISECONDS);
 
         this.currentChannel = builder.build();
+        this.confirmedChannel = false;
         this.previous = addr;
     }
 
@@ -108,5 +119,6 @@ public void shutdown() {
 
     public void clear() {
         this.previous = null;
+        this.confirmedChannel = false;
     }
 }

db-client-java/src/main/java/com/eventstore/dbclient/EventStoreDBClientBase.java

@@ -16,7 +16,7 @@ public class EventStoreDBClientBase {
     EventStoreDBClientBase(EventStoreDBClientSettings settings) {
         Discovery discovery;
 
-        if (settings.getHosts().length == 1) {
+        if (settings.getHosts().length == 1 && !settings.isDnsDiscover()) {
             discovery = new SingleNodeDiscovery(settings.getHosts()[0]);
         } else {
             discovery = new ClusterDiscovery(settings);

db-client-java/src/test/java/com/eventstore/dbclient/ClientTracker.java

@@ -74,6 +74,20 @@ public synchronized EventStoreDBClient getDefaultClient(Database database) {
                             continue;
                         }
 
+                        // In some rare occasions, it's possible for GHA to take much more time setting up a cluster
+                        // through docker compose. In this case, we recreate a fresh client in the case we exhausted
+                        // all discovery attempts and the connection got closed.
+                        if (e.getCause() instanceof ConnectionShutdownException && (settings.isDnsDiscover() || settings.getHosts().length > 1)) {
+                            logger.debug("Seems we exhausted all discovery attempts. Unusual but maybe docker is slow");
+                            try {
+                                Thread.sleep(500);
+                            } catch (InterruptedException ex) {
+                                throw new RuntimeException(ex);
+                            }
+                            defaultClient = EventStoreDBClient.create(settings);
+                            continue;
+                        }
+
                         throw new RuntimeException(e);
                     }
                 }

db-client-java/src/test/java/com/eventstore/dbclient/databases/ExternallyCreatedCluster.java

@@ -15,9 +15,8 @@ public ExternallyCreatedCluster(boolean secure) {
     public ConnectionSettingsBuilder defaultSettingsBuilder() {
         return EventStoreDBClientSettings
                 .builder()
+                .dnsDiscover(true)
                 .defaultCredentials("admin", "changeit")
-                .addHost("localhost", 2_111)
-                .addHost("localhost", 2_112)
                 .addHost("localhost", 2_113)
                 .tls(secure)
                 .tlsVerifyCert(false)

db-client-java/src/test/java/com/eventstore/dbclient/streams/DeadlineTests.java

@@ -11,7 +11,7 @@
 import java.util.concurrent.ExecutionException;
 
 public interface DeadlineTests extends ConnectionAware {
-    @Test
+    @RetryingTest(10)
      default void testDefaultDeadline() throws Throwable {
         EventStoreDBClient client = getDatabase().connectWith(opts ->
                 opts.defaultDeadline(1)

docker-compose.yml

@@ -1,106 +1,98 @@
-version: '3.5'
+version: "3.5"
 
 services:
   volumes-provisioner:
-    image: "hasnat/volumes-provisioner"
+    image: hasnat/volumes-provisioner
     environment:
       PROVISION_DIRECTORIES: "1000:1000:0755:/tmp/certs"
     volumes:
       - "./certs:/tmp/certs"
-    network_mode: "none"
+    network_mode: none
 
-  setup:
+  cert-gen:
     image: eventstore/es-gencert-cli:1.0.2
     entrypoint: bash
     user: "1000:1000"
     command: >
       -c "mkdir -p ./certs && cd /certs
       && es-gencert-cli create-ca
-      && es-gencert-cli create-node -out ./node1 --dns-names node1.eventstore
-      && es-gencert-cli create-node -out ./node2 --dns-names node2.eventstore
-      && es-gencert-cli create-node -out ./node3 --dns-names node3.eventstore
+      && es-gencert-cli create-node -out ./node1 -ip-addresses 127.0.0.1,172.30.240.11 -dns-names localhost
+      && es-gencert-cli create-node -out ./node2 -ip-addresses 127.0.0.1,172.30.240.12 -dns-names localhost
+      && es-gencert-cli create-node -out ./node3 -ip-addresses 127.0.0.1,172.30.240.13 -dns-names localhost
       && find . -type f -print0 | xargs -0 chmod 666"
-    container_name: setup
     volumes:
-      - ./certs:/certs
+      - "./certs:/certs"
     depends_on:
       - volumes-provisioner
 
-  node1.eventstore: &template
+  esdb-node1:
     image: ghcr.io/eventstore/eventstore:${CONTAINER_IMAGE_VERSION:-latest}
-    container_name: node1.eventstore
     env_file:
       - vars.env
     environment:
-      - EVENTSTORE_EXT_HOST_ADVERTISE_AS=node1.eventstore
-      - EVENTSTORE_INT_HOST_ADVERTISE_AS=node1.eventstore
-      - EVENTSTORE_GOSSIP_SEED=node2.eventstore:2113,node3.eventstore:2113
-      - EVENTSTORE_CERTIFICATE_FILE=/certs/node1/node.crt
-      - EVENTSTORE_CERTIFICATE_PRIVATE_KEY_FILE=/certs/node1/node.key
+      - EVENTSTORE_GOSSIP_SEED=172.30.240.12:2113,172.30.240.13:2113
+      - EVENTSTORE_INT_IP=172.30.240.11
+      - EVENTSTORE_CERTIFICATE_FILE=/etc/eventstore/certs/node1/node.crt
+      - EVENTSTORE_CERTIFICATE_PRIVATE_KEY_FILE=/etc/eventstore/certs/node1/node.key
       - EVENTSTORE_ADVERTISE_HTTP_PORT_TO_CLIENT_AS=2111
-      - EVENTSTORE_ADVERTISE_TCP_PORT_TO_CLIENT_AS=1111
-    healthcheck:
-      test:
-        [
-            'CMD-SHELL',
-            'curl --fail --insecure https://node1.eventstore:2113/health/live || exit 1',
-        ]
-      interval: 5s
-      timeout: 5s
-      retries: 24
     ports:
-      - 1111:1113
       - 2111:2113
+    networks:
+      clusternetwork:
+        ipv4_address: 172.30.240.11
     volumes:
-      - ./certs:/certs
+      - ./certs:/etc/eventstore/certs
+    restart: unless-stopped
     depends_on:
-      - setup
-    restart: always
+      - cert-gen
 
-  node2.eventstore:
-    <<: *template
-    container_name: node2.eventstore
+  esdb-node2:
+    image: ghcr.io/eventstore/eventstore:${CONTAINER_IMAGE_VERSION:-latest}
+    env_file:
+      - vars.env
     environment:
-      - EVENTSTORE_EXT_HOST_ADVERTISE_AS=node2.eventstore
-      - EVENTSTORE_INT_HOST_ADVERTISE_AS=node2.eventstore
-      - EVENTSTORE_GOSSIP_SEED=node1.eventstore:2113,node3.eventstore:2113
-      - EVENTSTORE_CERTIFICATE_FILE=/certs/node2/node.crt
-      - EVENTSTORE_CERTIFICATE_PRIVATE_KEY_FILE=/certs/node2/node.key
+      - EVENTSTORE_GOSSIP_SEED=172.30.240.11:2113,172.30.240.13:2113
+      - EVENTSTORE_INT_IP=172.30.240.12
+      - EVENTSTORE_CERTIFICATE_FILE=/etc/eventstore/certs/node2/node.crt
+      - EVENTSTORE_CERTIFICATE_PRIVATE_KEY_FILE=/etc/eventstore/certs/node2/node.key
       - EVENTSTORE_ADVERTISE_HTTP_PORT_TO_CLIENT_AS=2112
-      - EVENTSTORE_ADVERTISE_TCP_PORT_TO_CLIENT_AS=1112
-    healthcheck:
-      test:
-        [
-            'CMD-SHELL',
-            'curl --fail --insecure https://node2.eventstore:2113/health/live || exit 1',
-        ]
-      interval: 5s
-      timeout: 5s
-      retries: 24
     ports:
-      - 1112:1113
       - 2112:2113
+    networks:
+      clusternetwork:
+        ipv4_address: 172.30.240.12
+    volumes:
+      - ./certs:/etc/eventstore/certs
+    restart: unless-stopped
+    depends_on:
+      - cert-gen
 
-  node3.eventstore:
-    <<: *template
-    container_name: node3.eventstore
+  esdb-node3:
+    image: ghcr.io/eventstore/eventstore:${CONTAINER_IMAGE_VERSION:-latest}
+    env_file:
+      - vars.env
     environment:
-      - EVENTSTORE_EXT_HOST_ADVERTISE_AS=node3.eventstore
-      - EVENTSTORE_INT_HOST_ADVERTISE_AS=node3.eventstore
-      - EVENTSTORE_GOSSIP_SEED=node1.eventstore:2113,node2.eventstore:2113
-      - EVENTSTORE_CERTIFICATE_FILE=/certs/node3/node.crt
-      - EVENTSTORE_CERTIFICATE_PRIVATE_KEY_FILE=/certs/node3/node.key
+      - EVENTSTORE_GOSSIP_SEED=172.30.240.11:2113,172.30.240.12:2113
+      - EVENTSTORE_INT_IP=172.30.240.13
+      - EVENTSTORE_CERTIFICATE_FILE=/etc/eventstore/certs/node3/node.crt
+      - EVENTSTORE_CERTIFICATE_PRIVATE_KEY_FILE=/etc/eventstore/certs/node3/node.key
       - EVENTSTORE_ADVERTISE_HTTP_PORT_TO_CLIENT_AS=2113
-      - EVENTSTORE_ADVERTISE_TCP_PORT_TO_CLIENT_AS=1113
-    healthcheck:
-      test:
-        [
-            'CMD-SHELL',
-            'curl --fail --insecure https://node3.eventstore:2113/health/live || exit 1',
-        ]
-      interval: 5s
-      timeout: 5s
-      retries: 24
     ports:
-      - 1113:1113
       - 2113:2113
+    networks:
+      clusternetwork:
+        ipv4_address: 172.30.240.13
+    volumes:
+      - ./certs:/etc/eventstore/certs
+    restart: unless-stopped
+    depends_on:
+      - cert-gen
+
+networks:
+  clusternetwork:
+    name: eventstoredb.local
+    driver: bridge
+    ipam:
+      driver: default
+      config:
+        - subnet: 172.30.240.0/24

vars.env

@@ -1,7 +1,8 @@
 EVENTSTORE_CLUSTER_SIZE=3
 EVENTSTORE_RUN_PROJECTIONS=All
-EVENTSTORE_TRUSTED_ROOT_CERTIFICATES_PATH=/certs/ca
+EVENTSTORE_INT_TCP_PORT=1112
+EVENTSTORE_HTTP_PORT=2113
+EVENTSTORE_TRUSTED_ROOT_CERTIFICATES_PATH=/etc/eventstore/certs/ca
 EVENTSTORE_DISCOVER_VIA_DNS=false
-EVENTSTORE_ENABLE_EXTERNAL_TCP=true
 EVENTSTORE_ENABLE_ATOM_PUB_OVER_HTTP=true
-EVENTSTORE_ADVERTISE_HOST_TO_CLIENT_AS=localhost
+EVENTSTORE_ADVERTISE_HOST_TO_CLIENT_AS=localhost