Fix neo4j: container doesn't start currently (#264)

* Adjust Neo4j memory settings to avoid startup fail

* Upgrade Neo4j to community 5.26.0

* Add Neo4j indexes

templates/stores/neo4j.jsonnet

@@ -14,12 +14,14 @@ local images = import "values/images.jsonnet";
                     .with_image(images.neo4j)
                     .with_environment({
                         NEO4J_AUTH: "neo4j/password",
+                        NEO4J_server_memory_pagecache_size: "512m",
+                        NEO4J_server_memory_heap_max__size: "512m",
         //		NEO4J_server_bolt_listen__address: "0.0.0.0:7687",
         //		NEO4J_server_default__listen__address: "0.0.0.0",
         //		NEO4J_server_http_listen__address: "0.0.0.0:7474",
                     })
-                    .with_limits("1.0", "768M")
+                    .with_limits("1.0", "1536M")
-                    .with_reservations("0.5", "768M")
+                    .with_reservations("0.5", "1536M")
                     .with_port(7474, 7474, "api")
                     .with_port(7687, 7687, "api2")
                     .with_volume_mount(vol, "/data");

templates/values/images.jsonnet

@@ -1,7 +1,7 @@
 local version = import "version.jsonnet";
 {
     cassandra: "docker.io/cassandra:4.1.6",
-    neo4j: "docker.io/neo4j:5.22.0-community-bullseye",
+    neo4j: "docker.io/neo4j:5.26.0-community-bullseye",
     pulsar: "docker.io/apachepulsar/pulsar:3.3.1",
     pulsar_manager: "docker.io/apachepulsar/pulsar-manager:v0.4.0",
     etcd: "quay.io/coreos/etcd:v3.5.15",

trustgraph-flow/trustgraph/storage/triples/memgraph/write.py

@@ -55,6 +55,14 @@ class Processor(Consumer):
 
     def create_indexes(self, session):
 
+        # Race condition, index creation failure is ignored.  Right thing
+        # to do if the index already exists.  Wrong thing to do if it's
+        # because the store is not up yet
+
+        # In real-world cases, Memgraph will start up quicker than Pulsar
+        # and this process will restart several times until Pulsar arrives,
+        # so should be safe
+
         print("Create indexes...", flush=True)
 
         try:

trustgraph-flow/trustgraph/storage/triples/neo4j/write.py

@@ -50,6 +50,50 @@ class Processor(Consumer):
 
         self.io = GraphDatabase.driver(graph_host, auth=(username, password))
 
+        with self.io.session(database=self.db) as session:
+            self.create_indexes(session)
+
+    def create_indexes(self, session):
+
+        # Race condition, index creation failure is ignored.  Right thing
+        # to do if the index already exists.  Wrong thing to do if it's
+        # because the store is not up yet
+
+        # In real-world cases, Neo4j will start up quicker than Pulsar
+        # and this process will restart several times until Pulsar arrives,
+        # so should be safe
+
+        print("Create indexes...", flush=True)
+
+        try:
+            session.run(
+                "CREATE INDEX Node_uri FOR (n:Node) ON (n.uri)",
+            )
+        except Exception as e:
+            print(e, flush=True)
+            # Maybe index already exists
+            print("Index create failure ignored", flush=True)
+
+        try:
+            session.run(
+                "CREATE INDEX Literal_value FOR (n:Literal) ON (n.value)",
+            )
+        except Exception as e:
+            print(e, flush=True)
+            # Maybe index already exists
+            print("Index create failure ignored", flush=True)
+
+        try:
+            session.run(
+                "CREATE INDEX Rel_uri FOR ()-[r:Rel]-() ON (r.uri)",
+            )
+        except Exception as e:
+            print(e, flush=True)
+            # Maybe index already exists
+            print("Index create failure ignored", flush=True)
+
+        print("Index creation done", flush=True)
+
     def create_node(self, uri):
 
         print("Create node", uri)