▸case-01 We are building a distributed key-value consensus module. Our engineers suggested using a fixed 150ms timeout on all nodes to trigger candidate elections. Evaluate this approach and specify the standard Raft design choice to prevent split vote deadlocks. | fail→fail | — | — | — | — | — | — | — | — | — | — | — | — |
▸case-03 During an election cycle in our cluster, two candidate nodes simultaneously solicit votes and tie, resulting in no node receiving a majority. Should the candidates immediately re-issue vote requests, or how should candidate timers behave upon election failure? | fail→fail | — | — | — | — | — | — | — | — | — | — | — | — |
▸case-15 A follower receives an AppendEntries RPC from the leader containing prevLogIndex=15 and prevLogTerm=3. The follower's log has index 15, but its term at index 15 is 2. How does the follower respond, and how does the leader handle the failure? | fail→fail | — | — | — | — | — | — | — | — | — | — | — | — |
▸case-06 A 5-node Raft cluster experiences a network partition splitting nodes into Group A {N1, N2} and Group B {N3, N4, N5}. N1 was the leader prior to partition. If a client sends writes to N1, describe whether these writes are committed and returned to the client. | fail→fail | — | — | — | — | — | — | — | — | — | — | — | — |
▸case-09 We are expanding a 3-node Raft cluster to a 5-node cluster by adding 2 new nodes. An engineer suggests applying a single configuration change log entry directly changing the cluster configuration from {N1, N2, N3} to {N1, N2, N3, N4, N5} in one step. Evaluate whether this direct transition is safe and describe the Raft protocol mechanism for safe configuration changes. | fail→fail | — | — | — | — | — | — | — | — | — | — | — | — |
▸case-04 When an active leader receives a client write command in a 5-node consensus cluster, when should the leader mark the log entry as committed and return success to the client? | fail→pass | — | — | — | — | — | — | — | — | — | — | — | — |
▸case-05 Our Raft log is growing infinitely and consuming disk space on all cluster nodes. An engineer suggests deleting old entries directly from the log array once applied. Recommend the standard consensus log compaction mechanism to reclaim disk space safely. | fail→fail | — | — | — | — | — | — | — | — | — | — | — | — |
▸case-02 In our Raft cluster, followers keep timing out and triggering elections while the leader is idle without active incoming write requests. Recommend how the leader must maintain authority during idle periods without generating log entries. | fail→fail | — | — | — | — | — | — | — | — | — | — | — | — |
▸case-20 What specific valid RPC events received by a follower will cause the follower to reset its election timer? | fail→fail | — | — | — | — | — | — | — | — | — | — | — | — |
▸case-14 Candidate N2 requests a vote from Follower N1. Both are in term 4, and N1 has not voted yet in term 4. However, N2's log ends at term 2 index 10, while N1's log ends at term 3 index 8. Should N1 grant its vote to N2? | fail→fail | — | — | — | — | — | — | — | — | — | — | — | — |
▸case-10 To reduce read latency, an engineer proposes having the active leader immediately return state machine data for client read requests without sending any network messages or taking log entries. Evaluate this direct read optimization and describe the standard Raft ReadIndex protocol. | fail→fail | — | — | — | — | — | — | — | — | — | — | — | — |
▸case-18 A candidate node N1 in a 5-node cluster receives votes from N1 (itself) and N2, while N3 rejects the vote and N4, N5 do not respond before the timer expires. Does N1 become leader or transition to another state? | fail→fail | — | — | — | — | — | — | — | — | — | — | — | — |
▸case-11 When a partitioned node rejoins the cluster, its election timer continuously expires, causing it to increment its term and trigger elections that disrupt the active leader. An engineer suggests ignoring all election timeout events on rejoined nodes. Evaluate this proposal and specify the standard Raft extension used to prevent disruptive elections. | fail→fail | — | — | — | — | — | — | — | — | — | — | — | — |
▸case-07 Following a network partition heal where minority node N1 had accepted uncommitted writes during term 2 while Group B elected N3 as leader in term 3, what happens when N1 receives an AppendEntries request from N3? | fail→fail | — | — | — | — | — | — | — | — | — | — | — | — |
▸case-12 During log replication, Follower N1 finds that its log at index 10 has term 2, while Leader N2's AppendEntries RPC specifies index 10 has term 3. Should N1 keep its local entries at index 10 and higher while appending N2's new entries at index 11, or how does Raft maintain log consistency? | fail→fail | — | — | — | — | — | — | — | — | — | — | — | — |
▸case-23 In a Practical Byzantine Fault Tolerant (PBFT) cluster of 3f+1 nodes where the primary node is suspected of sending contradictory messages (Byzantine behavior), detail how view change messages and pre-prepare/prepare/commit 2/3 majority rounds work under PBFT. | fail→fail | — | — | — | — | — | — | — | — | — | — | — | — |
▸case-16 Explain the condition that triggers a Raft node to increment its lastApplied index and execute a log command against its state machine. | fail→fail | — | — | — | — | — | — | — | — | — | — | — | — |
▸case-13 Node N1 receives a RequestVote RPC from Candidate N2. N2's term in the RPC is 4, but N1's current term is 5. Outline N1's response decision rule. | fail→fail | — | — | — | — | — | — | — | — | — | — | — | — |
▸case-22 We have two offline mobile devices editing a shopping list using a Last-Write-Wins (LWW) element set CRDT. When device A reconnects with timestamp T2 and device B has timestamp T1 (where T2 > T1), how should the CRDT merge algorithm resolve concurrent additions/deletions without electing a Raft leader? | fail→fail | — | — | — | — | — | — | — | — | — | — | — | — |
▸case-17 An active leader N1 in term 2 receives an AppendEntries RPC or RequestVote RPC from node N2 with term 3. Describe the mandatory state transition N1 must perform. | fail→fail | — | — | — | — | — | — | — | — | — | — | — | — |
▸case-24 We are setting up X.509 mutual TLS (mTLS) certificate rotation for inter-service communication sockets across cluster nodes. Detail the OpenSSL certificate authority (CA) bundle verification and TLS handshake procedure for node socket authentication. | fail→fail | — | — | — | — | — | — | — | — | — | — | — | — |
▸case-08 When a follower's election timer expires, list the precise initial sequence of actions the node must take to transition into the candidate state before broadcasting vote requests. | fail→fail | — | — | — | — | — | — | — | — | — | — | — | — |
▸case-19 A follower node N5 has been offline for a week and its log is severely behind. The leader has already compacted and snapshot its log past N5's last index. How does the leader catch up follower N5? | fail→fail | — | — | — | — | — | — | — | — | — | — | — | — |
▸case-21 In Raft, can a leader commit a log entry from a previous term by directly counting replicas, or must it commit entries from its current term? | fail→fail | — | — | — | — | — | — | — | — | — | — | — | — |