BLACKHOLE_PEER_PORT_TX_RX_ALL drops all outgoing/incoming packets from/to the peer ports on all nodes, thus making cluster totally inoperable.

BLACKHOLE_PEER_PORT_TX_RX_LEADER drops all outgoing/incoming packets from/to the peer port on the active leader (isolated), and waits for "delay-ms" until recovery, in order to trigger election timeout.

BLACKHOLE_PEER_PORT_TX_RX_LEADER_UNTIL_TRIGGER_SNAPSHOT drops all outgoing/incoming packets from/to the peer port on the active leader, and waits for most up-to-date node (leader) applies the snapshot count of entries since the blackhole operation.

BLACKHOLE_PEER_PORT_TX_RX_ONE_FOLLOWER drops all outgoing/incoming packets from/to the peer port on a randomly chosen follower (non-leader), and waits for "delay-ms" until recovery.

BLACKHOLE_PEER_PORT_TX_RX_ONE_FOLLOWER_UNTIL_TRIGGER_SNAPSHOT drops all outgoing/incoming packets from/to the peer port on a randomly chosen follower (non-leader), and waits for most up-to-date node (leader) applies the snapshot count of entries since the blackhole operation.

BLACKHOLE_PEER_PORT_TX_RX_QUORUM drops all outgoing/incoming packets from/to the peer ports on majority nodes of cluster, thus losing its leader and cluster being inoperable.

DELAY_PEER_PORT_TX_RX_ALL delays outgoing/incoming packets from/to the peer ports on all nodes.

DELAY_PEER_PORT_TX_RX_LEADER delays outgoing/incoming packets from/to the peer port on the active leader.

DELAY_PEER_PORT_TX_RX_LEADER_UNTIL_TRIGGER_SNAPSHOT delays outgoing/incoming packets from/to the peer port on the active leader, and waits for most up-to-date node (current or new leader) applies the snapshot count of entries since the delay operation.

DELAY_PEER_PORT_TX_RX_ONE_FOLLOWER delays outgoing/incoming packets from/to the peer port on a randomly chosen follower (non-leader).

DELAY_PEER_PORT_TX_RX_ONE_FOLLOWER_UNTIL_TRIGGER_SNAPSHOT delays outgoing/incoming packets from/to the peer port on a randomly chosen follower (non-leader), and waits for most up-to-date node (leader) applies the snapshot count of entries since the delay operation.

DELAY_PEER_PORT_TX_RX_QUORUM delays outgoing/incoming packets from/to the peer ports on majority nodes of cluster.

EXTERNAL runs external failure injection scripts.

FAILPOINTS injects failpoints to etcd server runtime, triggering panics in critical code paths.

FAILPOINTS_WITH_DISK_IO_LATENCY injects high disk I/O latency failure in raftAfterSave code paths.

NO_FAIL_WITH_NO_STRESS_FOR_LIVENESS neither injects failures nor sends stressig client requests to the cluster, for the duration of "delay-ms".

NO_FAIL_WITH_STRESS stops injecting failures while testing the consistency and correctness under pressure loads, for the duration of "delay-ms".

RANDOM_DELAY_PEER_PORT_TX_RX_ALL delays outgoing/incoming packets from/to the peer ports on all nodes, with randomized time durations.

RANDOM_DELAY_PEER_PORT_TX_RX_LEADER delays outgoing/incoming packets from/to the peer port on the active leader with a randomized time duration.

RANDOM_DELAY_PEER_PORT_TX_RX_LEADER_UNTIL_TRIGGER_SNAPSHOT delays outgoing/incoming packets from/to the peer port on the active leader, with a randomized time duration.

RANDOM_DELAY_PEER_PORT_TX_RX_ONE_FOLLOWER delays outgoing/incoming packets from/to the peer port on a randomly chosen follower (non-leader) with a randomized time duration (thus isolated).

RANDOM_DELAY_PEER_PORT_TX_RX_ONE_FOLLOWER_UNTIL_TRIGGER_SNAPSHOT delays outgoing/incoming packets from/to the peer port on a randomly chosen follower (non-leader) with a randomized time duration, and waits for most up-to-date node (leader) applies the snapshot count of entries since the delay operation.

RANDOM_DELAY_PEER_PORT_TX_RX_QUORUM delays outgoing/incoming packets from/to the peer ports on majority nodes of cluster, with randomized time durations.

SIGQUIT_AND_REMOVE_LEADER stops the active leader node, deletes its data directories on disk, and removes this member from cluster.

SIGQUIT_AND_REMOVE_LEADER_UNTIL_TRIGGER_SNAPSHOT stops the active leader, deletes its data directories on disk, and removes this member from cluster (membership reconfiguration).

SIGQUIT_AND_REMOVE_ONE_FOLLOWER stops a randomly chosen follower (non-leader), deletes its data directories on disk, and removes this member from cluster (membership reconfiguration).

SIGQUIT_AND_REMOVE_ONE_FOLLOWER_UNTIL_TRIGGER_SNAPSHOT stops a randomly chosen follower, deletes its data directories on disk, and removes this member from cluster (membership reconfiguration).

SIGQUIT_AND_REMOVE_QUORUM_AND_RESTORE_LEADER_SNAPSHOT_FROM_SCRATCH first stops majority number of nodes, deletes data directories on those quorum nodes, to make the whole cluster inoperable.

SIGTERM_ALL stops the whole cluster but does not delete data directories on disk for next restart.

SIGTERM_LEADER stops the active leader node but does not delete its data directories on disk for next restart.

SIGTERM_LEADER_UNTIL_TRIGGER_SNAPSHOT stops the active leader node but does not delete its data directories on disk for next restart.

SIGTERM_ONE_FOLLOWER stops a randomly chosen follower (non-leader) but does not delete its data directories on disk for next restart.

SIGTERM_ONE_FOLLOWER_UNTIL_TRIGGER_SNAPSHOT stops a randomly chosen follower but does not delete its data directories on disk for next restart.

SIGTERM_QUORUM stops majority number of nodes to make the whole cluster inoperable but does not delete data directories on stopped nodes for next restart.

BLACKHOLE_PEER_PORT_TX_RX drops all outgoing/incoming packets from/to the peer port on target member's peer port.

DELAY_PEER_PORT_TX_RX delays all outgoing/incoming packets from/to the peer port on target member's peer port.

INITIAL_START_ETCD is only called to start etcd, the very first time.

NOT_STARTED is the agent status before etcd first start.

RESTART_ETCD is sent to restart killed etcd.

RESTART_FROM_SNAPSHOT is sent to trigger local member to restart and join an existing cluster that has been recovered from a snapshot.

RESTORE_RESTART_FROM_SNAPSHOT is sent to trigger local member to restore a cluster from existing snapshot from disk, and restart an etcd instance from recovered data.

SAVE_SNAPSHOT is sent to trigger local member to download its snapshot onto its local disk with the specified path from tester.

SIGQUIT_ETCD_AND_ARCHIVE_DATA is sent when consistency check failed, thus need to archive etcd data directories.

SIGQUIT_ETCD_AND_REMOVE_DATA kills etcd process and removes all data directories to simulate destroying the whole machine.

SIGQUIT_ETCD_AND_REMOVE_DATA_AND_STOP_AGENT destroys etcd process, etcd data, and agent server.

SIGTERM_ETCD pauses etcd process while keeping data directories and previous etcd configurations.

UNBLACKHOLE_PEER_PORT_TX_RX removes outgoing/incoming packet dropping.

UNDELAY_PEER_PORT_TX_RX removes all outgoing/incoming delays.

SnapshotInfo contains SAVE_SNAPSHOT request results.

UnimplementedTransportServer can be embedded to have forward compatible implementations.

TransportClient is the client API for Transport service.

TransportServer is the server API for Transport service.

Case defines various system faults or test case in distributed systems, in order to verify correct behavior of etcd servers and clients.