APPLYING CAP THEOREM TO BUILD DISTRIBUTED ROBUST PHOENIX API APP

PHOENIX API APP BACKED BY
COCKROACHDB

Transcript

1. APPLYING CAP TO BUILD DISTRIBUTED ROBUST PHOENIX API APP PHOENIX API APP BACKED BY COCKROACHDB
2. 2 Who Are we yeongsheng@odd-e.com
3. AGENDA ‣ YOUR WEAKEST LINK ‣ #NEWSQL + D.A.S.H ‣ COCKROACHDB ‣ WHEN RUBBER HITS THE ROAD ‣ SHOW ME THE CODE ‣ Q&A
4. APPLYING CAP TO BUILD DISTRIBUTED ROBUST PHOENIX API APP YOUR WEAKEST LINK ▸ Traditional RDBMS ▸ NoSQL ▸ NewSQL ▸ YugaqByteDB ▸ TiDB ▸ CockroachDB
5. APPLYING CAP TO BUILD DISTRIBUTED ROBUST PHOENIX API APP D.A.S.H ▸ Beyond 12-Factor-App ▸ 4 properties of CloudNative Ready DB DISPOSABILITY API SYMMETRY SHARED NOTHING HORIZONTAL SCALING D : A : S : H :
6. WHY? COCKROACHDB
7. APPLYING CAP TO BUILD DISTRIBUTED ROBUST PHOENIX API APP
8. APPLYING CAP TO BUILD DISTRIBUTED ROBUST PHOENIX API APP ENABLES GLOBAL, SCALABLE, RESILIENT SERVICES ▸ Make life easier for humans. Low-touch and highly automated distributed DB for operators yet simple to reason about for developers. ▸ Industry-leading consistency, even on massively scaled deployments. Enabling distributed transactions, yet removing the pain of eventual consistency issues. ▸ Always-on database that accepts reads and writes on all nodes without generating conflicts. ▸ Flexible deployment in any environment, without tying you to any platform or vendor. ▸ Support familiar tools for working with relational data (i.e., SQL).
9. WHAT GIVES? CONSISTENCY
10. APPLYING CAP TO BUILD DISTRIBUTED ROBUST PHOENIX API APP PERILS OF INCONSISTENCIES ▸ Emin Gün Sirer, an Associate Professor at Cornell University, wrote a blog post blaming eventually consistent data stores for the lost bitcoins. He mentions MongoDB, Cassandra and Riak among the NoSQL solutions that are vulnerable to banking thefts because:
11. APPLYING CAP TO BUILD DISTRIBUTED ROBUST PHOENIX API APP CONSENSUS REPLICATION ▸ Raft ▸ Commit when quorum has written data (nodes > = 3) ▸ 1 consensus group per data range ution replication protocol (Raft) transactions t, repair, rebalance Gossip / Raft SQL API Distributed, transactional KV Node 1 Node 2 Node 3
12. APPLYING CAP TO BUILD DISTRIBUTED ROBUST PHOENIX API APP CONSISTENCY<-CONSENSUS<-MULTI-ACTIVE AVAILABILITY ▸ “CONSISTENCY” - ACID semantics & CAP theorem. Data should be anomaly-free. ▸ “CLUSTER” - Single logical DB, multiple nodes, joined together to form a uniform consistent cluster. ▸ “RANGE” - All data (tables, indices, etc) as a giant sorted map of KV pairs; 64 MB in size per range; Auto split and balanced across nodes. ▸ “LEASEHOLDER” - Ranges replicated across nodes & 1 replica of range holds “range lease”; coordinates all reads/writes to that range.
13. APPLYING CAP TO BUILD DISTRIBUTED ROBUST PHOENIX API APP CONSISTENCY<-CONSENSUS<-MULTI-ACTIVE AVAILABILITY ▸ “RAFTLEADER” - 1 replica for each range is “leader” for writes; Leader coordinates all writes to raft group (followers); Quorum consensus for writes. Leaseholder usually also Raftleader. When a write doesn't achieve consensus, forward progress halts to maintain consistency within the cluster. Ensures ACID semantics for multi-tables data operations are consistent. ▸ “REPLICATION” - Synchronous distribution of copies of data that are ensured to be consistent. ▸ “MULTI-ACTIVE AVAILABILITY” - Each node in cluster can handle reads/ writes for a subset of stored data (range and quorum consensus concepts). Symmetrical -> works nicely with load balancers over SQL API (PostgreSQL wire format).
14. APPLYING CAP TO BUILD DISTRIBUTED ROBUST PHOENIX API APP @cockroachdb Put “cherry” Leader apricot banana blueberry cherry grape Follower apricot banana blueberry grape Follower apricot banana blueberry grape Consensus Replication Put “cherry” Leader apricot banana blueberry cherry grape Follower apricot banana blueberry grape Follower apricot banana blueberry grape Put “cherry” Consensus Replication RAFT CONSENSUS PROTOCOL
15. APPLYING CAP TO BUILD DISTRIBUTED ROBUST PHOENIX API APP @cockroach Write committed when 2 out of 3 nodes have written data Follower apricot banana blueberry cherry grape Follower apricot banana blueberry grape Consensus Replication Put “cherry” Leader apricot banana blueberry cherry grape Put “cherry”
16. APPLYING CAP TO BUILD DISTRIBUTED ROBUST PHOENIX API APP @cockroachdb Follower apricot banana blueberry cherry grape Follower apricot banana blueberry grape Consensus Replication Put “cherry” Leader apricot banana blueberry cherry grape Put “cherry” Ack Ack Follower apricot banana blueberry cherry grape Follower apricot banana blueberry grape Consensus Replication Put “cherry” Leader apricot banana blueberry cherry grape Put “cherry” cherry Ack
17. CAN IT BE TRUSTED? WHEN RUBBER HITS THE ROAD
18. APPLYING CAP TO BUILD DISTRIBUTED ROBUST PHOENIX API APP COCKROACHDB PASSES JENSEN TESTING ▸ CockroachDB Beta passes Jepsen testing ▸ Jepsen tests lessons 2+ years on... ▸ Register, Bank, Monotonic & Sets ▸ G2, Sequential, Comments & Nemesis modes (clock skews, network failures)
19. APPLYING CAP TO BUILD DISTRIBUTED ROBUST PHOENIX API APP
20. DEMO
21. APPLYING CAP TO BUILD DISTRIBUTED ROBUST PHOENIX API APP ▸ https://github.com/yeongsheng-tan/lunchbox_api.git ▸ Branches ▸ master ▸ cockroachdb_postgrex_cdb ▸ cockroachdb_postgrex_cdb_with_api_jwt_auth ▸ cockroachdb_postgrex_cdb_api_jwt_auth_e2e_tests
22. APPLYING CAP TO BUILD DISTRIBUTED ROBUST PHOENIX API APP WHAT IS NEEDED? ▸ Production code -> hex package: postgrex_cdb (yet to support correlated sub- queries) ▸ Test code -> hex package: ecto_replay_sandbox ▸ No prod code changes!!! ▸ Test code changes.
23. APPLYING CAP TO BUILD DISTRIBUTED ROBUST PHOENIX API APP
24. APPLYING CAP TO BUILD DISTRIBUTED ROBUST PHOENIX API APP CREDITS ▸ Nate Stewart ▸ Ben Darnell ▸ Aphyr Jepsen ▸ Diana Hsieh ▸ Emin Gün Sirer