Neo4j Drivers Best Practices
1 like177 views
The document discusses best practices for using Neo4j drivers. It recommends connecting with neo4j+s:// for security and routing context. Drivers should verify connectivity before queries and reuse a single driver instance. Transactions should use explicit functions and parameters to avoid overloading servers. Bookmarks provide causal consistency by allowing queries to read their own writes.
Neo4j Drivers Best Practices
- 1. © 2022 Neo4j, Inc. All rights reserved. Neo4j Drivers Best Practices M. David Allen Senior Director, Developer Relations mdavidallen These slides: https://dev.neo4j.com/neo4j-driver-best-practices
- 2. © 2022 Neo4j, Inc. All rights reserved. 2 Outline ● What's a driver and how does it work? ● What are the best practices? ● Common Pitfalls ● Understanding consistency in Neo4j
- 3. © 2022 Neo4j, Inc. All rights reserved. 3 What Drivers Do
- 4. © 2022 Neo4j, Inc. All rights reserved. 4 What Drivers Work Against
- 5. © 2022 Neo4j, Inc. All rights reserved. 5 Routing Tables
- 6. © 2022 Neo4j, Inc. All rights reserved. 6 What Routing Tables Look Like CALL dbms.cluster.routing.getRoutingTable({}) YIELD ttl, servers UNWIND servers as server RETURN ttl, server.role, server.addresses;
- 7. © 2022 Neo4j, Inc. All rights reserved. 7 Best Practices
- 8. © 2022 Neo4j, Inc. All rights reserved. 8 Connect with neo4j+s:// Whenever Possible You want routing context, and you want to be secured via full certificate validation, to prevent man in the middle attacks neo4j+s:// is the default in AuraDB, and going forward bolt is likely to be phased out.
- 9. © 2022 Neo4j, Inc. All rights reserved. 9 Use Very Recent Versions of Neo4j Drivers ● Introduction of AuraDB has introduced changes to connection details and routing details for clusters ● Newer drivers benefit from performance enhancements, and reduced connection errors in cloud environments
- 10. © 2022 Neo4j, Inc. All rights reserved. 10 Verify Connectivity Before Issuing Queries ● Push connection errors to code initialization time, not query time ● Pay one-time cost in a predictable place ● Drivers initialize a connection pool
- 11. © 2022 Neo4j, Inc. All rights reserved. 11 Create One Driver Instance and Hold Onto It Expensive: ● Drivers ● Connection Pools Cheap: ● Sessions ● Queries 11
- 12. © 2022 Neo4j, Inc. All rights reserved. 12 Where This Goes Wrong: What Not To Do ● AWS Lambda functions, serverless functions in general ● Users create a driver for every request, and don't verify connectivity ● Observed problems: ○ Serverless functions take too long to execute (paying connection pool cost over and over again) ○ Connection errors to the database fail at function invocation time, not startup
- 13. © 2022 Neo4j, Inc. All rights reserved. 13 Use Explicit Transaction Functions with driver.session as session: session.run("CREATE (p:Person { name: $name})", name="David") def add_person(self, name): with driver.session() as session: session.write_transaction(self.create_person_node, name) def create_person_node(tx, name): return tx.run("CREATE (a:Person {name: $name})", name=name)
- 14. © 2022 Neo4j, Inc. All rights reserved. 14 Use Query Parameters Whenever Possible def create_person_node(tx, name): tx.run("CREATE (a:Person {name: $name})", name=name ) def create_person_node(tx, name): tx.run("CREATE (a:Person {name: '%s'})" % name ) ● Avoids Cypher injection attack potential ● Reuses compiled queries in the database
- 15. © 2022 Neo4j, Inc. All rights reserved. 15 Where This Goes Wrong: Pitfalls ● "Autocommit transactions" (session.run()) always send traffic to the leader ● Heavy workloads overload the leader, leaving followers idle ● Without parameterized queries, database has to work harder to execute the same query over and over
- 16. © 2022 Neo4j, Inc. All rights reserved. 16 Process Database Results Within your Tx Function
- 17. © 2022 Neo4j, Inc. All rights reserved. 17 Returning Vanilla Python Objects
- 18. © 2022 Neo4j, Inc. All rights reserved. 18 Causal Consistency in Neo4j (part 1) CREATE (p:Person { name: "David" }) MATCH (p:Person) WHERE p.name = "David" RETURN count(p) 1st query succeeds when a majority of cluster members have acknowledged the write. Generally that will be 2 out of 3. What the second query returns depends on timing, and which cluster member you ask!
- 19. © 2022 Neo4j, Inc. All rights reserved. 19 Causal Consistency in Neo4j: Bookmarks (part 2) User: CREATE (p:Person { name: "David" }) Server: OK, got it. Bookmark XYZ. User: As of point in time XYZ, MATCH (p:Person) WHERE p.name = "David" RETURN count(p) Server: OK, cool. Result=1 With bookmarks, you always get what you expect, because you query "as of point in time" from your previous queries That is, you read your own writes In other words: you read the changes you cause (causal consistency)
- 20. © 2022 Neo4j, Inc. All rights reserved. 20 Causal Consistency (part 3): Bookmark chaining ● Session objects in Neo4j drivers chain bookmarks automatically ● As long as you reuse a session, you get the expected behavior for free ● If you create new sessions, you can still pass bookmarks manually
- 21. © 2022 Neo4j, Inc. All rights reserved. 21 Advanced Issues ● Distributed microservices that need to read each other's writes ○ Generally requires bookmark passing, for example via Redis
- 22. © 2022 Neo4j, Inc. All rights reserved. 22 Summary: Best Practices Make your Code: ● More performant (minimizing setup/connection time) ● More secure (no Cypher injection potential, encrypted communications) ● More robust & portable to different Neo4j environments (single instance vs. cluster)