Scenario catalog

This release has 135 scenarios in 16 modules, in the order the playground's library shows them. A scenario's number is its id. A title opens the scenario's page, which walks through it step by step; "open in the playground" starts it in the playground, paused on its walkthrough.

Anatomy

• Topic, partition, offset 00.1.1 · open in the playground
• One record, end to end 00.1.2 · open in the playground
• Keys and partitioning 00.1.3 · open in the playground
• Offsets are per-partition 00.1.4 · open in the playground
• Brokers, replicas, leader 00.1.5 · open in the playground
• LEO vs high watermark 00.2.1 · open in the playground
• Reading up to the HW 00.2.2 · open in the playground
• Log start offset 00.2.3 · open in the playground

Producers & EOS

• acks=0: fire and forget 01.1.1 · open in the playground
• acks=1: leader-only ack 01.1.2 · open in the playground
• acks trade-off 01.1.3 · open in the playground
• acks=1 durability gap 01.1.4 · open in the playground
• linger.ms and batch.size 01.2.1 · open in the playground
• Producer partitioners 01.2.2 · open in the playground
• Compression and batch shape 01.2.3 · open in the playground
• Backpressure: buffer.memory and max.block.ms 01.2.4 · open in the playground
• Idempotence retry 01.3.1 · open in the playground
• The duplicate problem 01.3.2 · open in the playground
• Idempotent dedup vs duplicate 01.3.3 · open in the playground
• Out-of-order detection and max.in.flight 01.3.4 · open in the playground

Consumer groups

• Subscribe, poll, fetch loop 02.1.1 · open in the playground
• auto.offset.reset 02.1.2 · open in the playground
• Committing offsets 02.1.3 · open in the playground
• Replay: rewind to an earlier offset 02.1.4 · open in the playground
• Partitions split across members 02.2.1 · open in the playground
• Add a consumer: rebalance 02.2.2 · open in the playground
• Assignors: range vs round-robin vs sticky 02.2.3 · open in the playground
• Rebalance: eager vs cooperative 02.2.4 · open in the playground
• Consumer crash & session timeout 02.3.1 · open in the playground
• Stuck consumer & max.poll.interval.ms 02.3.2 · open in the playground
• Static membership: avoid rebalance 02.3.3 · open in the playground
• Sustained consumer lag 02.3.4 · open in the playground
• Evicted-member ghost & survivor takeover 02.3.5 · open in the playground

Replication

• The ISR set 03.1.1 · open in the playground
• min.ISR boundary 03.1.2 · open in the playground
• Slow follower 03.1.3 · open in the playground
• HW recompute on ISR shrink 03.1.4 · open in the playground
• Canonical failure 03.2.1 · open in the playground
• Leader balancing 03.2.2 · open in the playground
• Unclean leader 03.2.3 · open in the playground
• Unclean election: permanent data loss 03.2.4 · open in the playground
• Returning replica and log truncation 03.2.5 · open in the playground
• Rack-aware placement 03.3.1 · open in the playground
• Rack rebalance 03.3.2 · open in the playground
• Change replication factor live 03.3.3 · open in the playground

Transactions

• At-most-once, at-least-once, exactly-once 04.1.1 · open in the playground
• Where duplicates come from 04.1.2 · open in the playground
• Transactional write: commit vs abort 04.2.1 · open in the playground
• Transactions & read_committed 04.2.2 · open in the playground
• Concurrent transactions & LSO 04.2.3 · open in the playground
• Transaction timeout 04.2.4 · open in the playground
• read_committed during a leader election 04.2.5 · open in the playground
• Consume-process-produce pipeline 04.3.1 · open in the playground
• EOS - offsets in a txn + fencing 04.3.2 · open in the playground
• exactly_once_v2 end-to-end 04.3.3 · open in the playground

Storage

• Time-based retention (retention.ms) 05.1.1 · open in the playground
• Size-based retention (retention.bytes) 05.1.2 · open in the playground
• Offset out of range 05.1.3 · open in the playground
• Compaction 05.2.1 · open in the playground
• Tombstones & delete.retention.ms 05.2.2 · open in the playground
• compact,delete policy 05.2.3 · open in the playground
• Tiered storage 05.3.1 · open in the playground
• Reading from the remote tier 05.3.2 · open in the playground
• Local vs total retention 05.3.3 · open in the playground
• Diskless: the write path 05.4.1 · open in the playground
• No replication across AZs 05.4.2 · open in the playground
• The price is latency 05.4.3 · open in the playground

KRaft

• The controller's job 06.0.1 · open in the playground
• KRaft quorum 06.0.2 · open in the playground
• Controller failover & metadata catch-up 06.0.3 · open in the playground
• ZooKeeper vs KRaft: the same cluster, two control planes 06.0.4 · open in the playground

Operations

• Scale out + rebalance 07.1.1 · open in the playground
• Scale in: drain and remove a broker 07.1.2 · open in the playground
• Partition reassignment by hand 07.1.3 · open in the playground
• Throttled reassignment under load 07.1.4 · open in the playground
• Live config changes 07.2.1 · open in the playground
• Quotas: throttling a producer 07.2.2 · open in the playground

Chaos

• Single broker hard kill 08.0.1 · open in the playground
• Network partition: split-brain 08.0.2 · open in the playground
• Minority vs majority 08.0.3 · open in the playground
• Cascading failure: losing the ISR one by one 08.0.4 · open in the playground
• Heal the partition: ISR re-expansion and catch-up 08.0.5 · open in the playground
• Correlated failure: a whole rack goes down 08.0.6 · open in the playground

Disaster recovery

• Stretched cluster across 2 DCs 09.1.1 · open in the playground
• Stretched 2.5-DC 09.1.2 · open in the playground
• Stretched 3-DC 09.1.3 · open in the playground
• 2-DC freeze 09.1.4 · open in the playground
• Losing a whole DC 09.1.5 · open in the playground
• Intra-DC partition split 09.1.6 · open in the playground
• Active-passive mirror 09.2.1 · open in the playground
• Failover + RPO 09.2.2 · open in the playground
• Active-active multi-cluster 09.2.3 · open in the playground
• Offset translation across clusters 09.2.4 · open in the playground
• Failback after the primary recovers 09.2.5 · open in the playground
• Active-active DR drill 09.2.6 · open in the playground
• MRC observers 10.1.1 · open in the playground
• Observer lag and async replication 10.1.2 · open in the playground
• Observer promotion 10.1.3 · open in the playground
• Observer promotion needs a reachable leader 10.1.4 · open in the playground
• Durability vs latency placement 10.1.5 · open in the playground
• Cluster Linking: byte-for-byte offset-preserving mirror 10.2.1 · open in the playground
• Mirror topics are read-only; promotion makes them writable 10.2.2 · open in the playground
• Cluster Linking vs MirrorMaker 2: when to use which 10.2.3 · open in the playground
• Consumer offset sync across a link 10.2.4 · open in the playground
• Bidirectional links and clean failback 10.2.5 · open in the playground

Tiered storage

• Confluent storage lifecycle: offload, read-back, expire 10.3.1 · open in the playground
• Infinite retention on the remote tier 10.3.2 · open in the playground

Schema registry

• Subjects and versions 10.4.1 · open in the playground
• Compatibility modes (backward/forward/full) 10.4.2 · open in the playground
• A breaking schema change is rejected 10.4.3 · open in the playground

Authorization (ACLs & RBAC)

• ACLs: authorizing a producer 10.5.1 · open in the playground
• RBAC: role bindings 10.5.2 · open in the playground

Share groups

• Share group queue 11.0.1 · open in the playground
• Cooperative consumption beyond partition count 11.0.2 · open in the playground
• Record acquisition locks and lock timeout 11.0.3 · open in the playground
• Acknowledge: accept, release, reject 11.0.4 · open in the playground
• Share group poison pill 11.0.5 · open in the playground
• Implicit vs explicit acknowledgement mode 11.0.6 · open in the playground
• Share group: batch_optimized acquire 11.0.7 · open in the playground
• RENEW: extending an acquisition lock 11.0.8 · open in the playground
• SPSO/SPEO/lag: the share-group state machine 11.0.9 · open in the playground

Capstones

• Capstone: exactly-once payments pipeline survives a region loss 12.0.1 · open in the playground
• Capstone: survive a regional outage with zero data loss 12.0.2 · open in the playground
• Capstone: migrate a cluster with Cluster Linking, zero downtime 12.0.3 · open in the playground
• Capstone: multi-DC active-active event backbone 12.0.4 · open in the playground
• Capstone: tune a high-throughput, low-latency pipeline 12.0.5 · open in the playground

Terminal ops

• Where am I: hosts, prompt, ssh 13.1.1 · open in the playground
• Diagnose lag before opening the inspector 13.1.2 · open in the playground
• A partition on disk, leader vs follower 13.1.3 · open in the playground
• Find the unhealthy partitions 13.1.4 · open in the playground
• Reset offsets the right way 13.2.1 · open in the playground
• Change a topic config live 13.2.2 · open in the playground
• Move a partition: execute, throttle, verify 13.2.3 · open in the playground
• Stop a broker from its own shell 13.2.4 · open in the playground
• Two clusters, two views 13.3.1 · open in the playground