Advanced Systems — Optional Laboratories

Infrastructure and adversarial experiments after the core curriculum.

Independent educational material; not endorsed by the named companies.

Track progress0 / 0

Progress is stored only in this browser.

1. Database concurrency and recovery

PostgreSQL Integration Lab: Concurrent Reservations and Crash RecoveryUse independent processes and a real PostgreSQL service to test conditional stock updates, idempotency and durable outbox commits.
PostgreSQL Serializable Lab: Write Skew and Whole-Transaction RetriesReproduce real PostgreSQL write skew at READ COMMITTED, block it with SERIALIZABLE and retry the full transaction.
SQLite Multi-Process Workshop: Atomic Reservations and Crash RecoveryTest real concurrent processes, file-backed SQLite transactions, stock constraints and replay after hard process termination.
PostgreSQL Physical Replication Lab: Standby Replay and Manual PromotionExercise real PostgreSQL 17 primary/standby replication using pg_basebackup, observe replay, stop the primary and promote the standby.
PostgreSQL Failover Safety: Fencing Receipts and Promotion GatesImplement and test fail-closed promotion decisions, reject stale fence evidence and examine real PostgreSQL standby promotion boundaries.
PostgreSQL Recovery Measurements: RPO, RTO and Network SplitMeasure controlled PostgreSQL promotion phases, observe a replayed commit and show why an isolated primary still needs fencing.

RabbitMQ Integration Lab: Redelivery and Transactional InboxTest a real RabbitMQ consumer disconnect, redelivered messages and PostgreSQL inbox deduplication with manual acknowledgments.
RabbitMQ Restart Lab: Durable Queues and Confirmed MessagesRestart a real single-node RabbitMQ container after a confirmed persistent publish and verify exact recovery, without claiming quorum failover.
RabbitMQ Quorum Leader Failover: Three-Node Executable LabTest real RabbitMQ quorum replication: publish with confirms, stop the initial leader, recover data, and confirm writes on the surviving majority.
RabbitMQ Quorum Majority Loss: Fail Closed and Recover WritesTest loss of a three-member queue majority, bounded publisher confirms, and successful writes after restoration on real RabbitMQ nodes.
RabbitMQ Live Network Partition: Running Brokers Without a QuorumDisconnect a running broker from its Docker network, observe minority confirmation behavior and restore the real quorum.

Concurrency Interview Workshop: Races, CAS and InterleavingsDerive a two-worker lost-update counterexample, fix it with versioned CAS, enumerate schedules and test thread synchronization.
Fault-Injection Workshop: Durable Outbox, Crashes and Duplicate DeliveryTest precommit and postcommit crashes, at-least-once outbox publication, receiver deduplication and recovery boundaries.
Property-Based Algorithm Testing: Oracles, Shrinking and Metamorphic ChecksGenerate adversarial cases with Hypothesis; compare binary search and signed-subarray algorithms against independent oracles.
Strongly Connected Components and Condensation GraphsDerive strongly connected components, prove why condensation is a DAG and implement iterative Kosaraju with complexity and edge-case checks.
Maximum Flow, Minimum Cuts and Bipartite MatchingDerive flow conservation, residual networks, augmenting paths, the max-flow min-cut theorem and bipartite matching reduction with tested code.
Observability: Metrics, Logs, Traces and SLOsModel end-to-end latency, golden signals, trace propagation, SLO error budgets and high-cardinality pitfalls with tested calculations.

Experiments only establish their stated lab conditions, not cross-host production fault tolerance or universally correct systems.