package slog.internal

Get desktop application:
View/edit binary Protocol Buffers messages

Used in: ForwardBatchData

• uint64 id = 1

• repeated Transaction transactions = 2

• TransactionType transaction_type = 3

  All txns in this batch has this same type. Can either be SINGLE_HOME or MULTI_HOME. LOCK_ONLY txns are also considered SINGLE_HOME.

• repeated TransactionEventInfo events = 4

  For recording the event time for all transactions in this batch simultaneously

Used in: Request

• uint64 batch_id = 1

Used in: Request

• uint64 tag = 1

• uint32 channel = 2

• bool stop = 3

* The schema of a configuration file.

• string protocol = 1

  Protocol for the zmq sockets in the broker. Use "tcp" for normal running and "icp" for unit and integration tests

• repeated Replica replicas = 2

  Replica groups. Each group has a list of machine addresses with the size equal to number of partitions

• repeated uint32 broker_ports = 3

  Port for the zmq sockets in the broker. Each value will create a new broker thread

• uint32 server_port = 4

  Port for the server

• uint32 forwarder_port = 5

  Port for the forwarder

• uint32 sequencer_port = 6

  Port for the sequencer

• uint32 num_partitions = 7

  Number of partitions per replica

• oneof partitioning

  Partitioning scheme

  • HashPartitioning hash_partitioning = 8

  • SimplePartitioning simple_partitioning = 9

  • TPCCPartitioning tpcc_partitioning = 10

• uint32 num_workers = 11

  Number of worker threads for processing the transactions

• uint64 mh_orderer_batch_duration = 12

  How long the mh orderer waits for batching

• uint64 forwarder_batch_duration = 13

  How long the forwarder waits for batching

• uint64 sequencer_batch_duration = 14

  How long the sequencer waits for batching

• int32 sequencer_batch_size = 15

  How big can a batch be in number of txns

• bool sequencer_rrr = 16

  By default, a sequencer sticks to a single partition of a region to send the batch to. If this option is true, the partitions are selected in a round-robin manner (rrr: round-robin replication).

• uint32 replication_factor = 17

  Number of replicas that need to be synchronously replicated to

• repeated string replication_order = 18

  Order of replicas for synchronous replication. This is a list of strings each of which corresponds to a replica. Each string is an ordered comma-separated list of replica ids that the current replica will synchronously replicate to based on replication order. The current replica will be ignored if exists in the list. For example, ['1,2', '2,1,0', ''] means that, with a replication factor of 3, replica 0 will synchronously replicate to 1 and 2, replica 1 will replicate to 2 and 0, and replica 2 will not synchronously replicate to anywhere.

• optional ReplicationDelayExperiment replication_delay = 19

  Replication of txn batches will be delayed to simulate uneven network latency

• repeated TransactionEvent enabled_events = 20

  Enable recording for the specified events

• bool bypass_mh_orderer = 21

  For multi-home txn, send lock-only txns directly to the regions, skipping the global orderer

• repeated CpuPinning cpu_pinnings = 22

  Pin each module to a cpu

• bool return_dummy_txn = 23

  Return dummy txns back to the client instead of full results

• int32 recv_retries = 24

  Number of retries after calling zmq::recv something

• ExecutionType execution_type = 25

  Way of executing code of a txn

• bool synchronized_batching = 26

  For each MH transaction, the involved regions will wait in such a way that their respective lock-only txns are put into the batches at approximately the same time

• uint32 sample_rate = 27

  Percentage of metrics to sample

• int32 broker_rcvbuf = 28

  Kernel receiving buffer size (bytes) of a broker socket

• int32 long_sender_sndbuf = 29

  Kernel sending buffer size (bytes) of long-distance sockets (e.g. those in the Forwarder and Sequencer)

Used in: Configuration

• ModuleId module = 1

• uint32 cpu = 2

• oneof type

  • Request request = 1

  • Response response = 2

  • bytes raw = 3

• uint32 from = 4

Used in: Configuration

• KEY_VALUE = 0

• NOOP = 1

• TPC_C = 2

Used in: Request

• optional Transaction txn = 1

• uint32 partition = 2

Used in: Request

• repeated Batch batch_data = 1

• uint32 home = 2

• uint32 home_position = 3

  Batches generated by the same machine need to follow the order of creation. This field is used to number the batches following that order. It always start from 0 and increment by 1

Used in: Request

• oneof locality

  • LocalBatchOrder local_batch_order = 1

  • RemoteBatchOrder remote_batch_order = 2

Used in: Request

• optional Transaction txn = 1

* With hash partitioning, each key is interpreted as a byte string. The keys are distributed to the partitions based on their hash values. Initial data can be loaded from an external file.

Used in: Configuration

• uint32 partition_key_num_bytes = 1

  Number of prefix bytes of a key to use for partitioning

Used in: LookupMasterResponse

• bytes key = 1

• optional MasterMetadata metadata = 2

Used in: ForwardBatchOrder

• uint32 queue_id = 1

  queue_id is the partition that the batch is generated from

• uint32 slot = 2

• uint32 leader = 3

Used in: Request

• repeated uint64 txn_ids = 1

• repeated bytes keys = 2

Used in: Response

• repeated uint64 txn_ids = 1

• repeated KeyMasterMetadata lookup_results = 2

Used in: Request

• uint32 ballot = 1

• uint32 slot = 2

• uint64 value = 3

Used in: Response

• uint32 ballot = 1

• uint32 slot = 2

Used in: Request

• uint32 slot = 1

• uint64 value = 2

• uint32 leader = 3

Used in: Response

• uint32 slot = 1

Used in: Request

• uint64 value = 1

Used in: Request

• int64 time = 1

• uint32 target = 2

• uint32 from_channel = 3

* For debugging and testing purposes

Used in: Response

• int64 time = 1

• uint32 target = 2

Used in: ForwardBatchOrder

• uint64 batch_id = 1

• uint32 slot = 2

• uint32 home = 3

• bool need_ack = 4

Used in: Request

• uint64 txn_id = 1

• uint32 partition = 2

• repeated KeyValueEntry reads = 3

• bool will_abort = 4

• string abort_reason = 5

Used in: Configuration

• repeated string addresses = 1

  List of all server addresses in the system. This list must have the size equal to number of partitions If protocol is "tcp", these are IP addresses. If protocol is "icp", these are filesystem paths.

• repeated string public_addresses = 2

  AWS public addresses for the servers. This field is only used by the admin tool. If not specified, the addresses field is used instead.

• repeated string client_addresses = 3

  List of machines used for benchmarking. This field is only used by the admin tool

• string distance_ranking = 4

  Comma-separated string of replica id (e.g "1,3,2") ordered from closest replica to furthest replica, not including this replica. This field is only used by the admin tool.

Used in: Configuration

• uint32 delay_pct = 1

  Percent of batches that are delayed

• uint32 delay_amount_ms = 2

  Average duration of delay in milliseconds

* The first message of a communication between two entities is always a Request. Some types of Request needs a Response (e.g. LookUpMasterRequest); names of these requests end with "Request". Some others are one-way requests (e.g. PaxosPropose) and do not end with "Request".

Used in: Envelope

• oneof type

  • Ping ping = 1

  • Signal signal = 2

  • BrokerRedirect broker_redirect = 3

  • ForwardTransaction forward_txn = 4

  • LookupMasterRequest lookup_master = 5

  • ForwardBatchData forward_batch_data = 6

  • ForwardBatchOrder forward_batch_order = 7

  • BatchReplicationAck batch_replication_ack = 8

  • PaxosPropose paxos_propose = 9

  • PaxosAcceptRequest paxos_accept = 10

  • PaxosCommitRequest paxos_commit = 11

  • RemoteReadResult remote_read_result = 12

  • FinishedSubtransaction finished_subtxn = 13

  • StatsRequest stats = 14

* A response is always preceeded by a Request

Used in: Envelope

• oneof type

  • Pong pong = 1

  • LookupMasterResponse lookup_master = 2

  • PaxosAcceptResponse paxos_accept = 3

  • PaxosCommitResponse paxos_commit = 4

  • StatsResponse stats = 6

* Generic signal message

Used in: Request

(message has no fields)

* With simple partitioning, each key is assumed to be a number. The keys are distributed to the partitions based on key values modulo the number of partitions. Initial data cannot be loaded from an external file but can be generated at startup. The following is an example of home and partition assignment of the generated data with 3 partitions and 4 regions. home | 0 1 2 3 0 1 2 3 0 ... ------------|------------------------------- partition 0 | 0 3 6 9 12 15 18 21 24 ... partition 1 | 1 4 7 10 13 16 19 22 25 ... partition 2 | 2 5 8 11 14 17 20 23 26 ... ------------|------------------------------- | keys

Used in: Configuration

• uint64 num_records = 1

  Number of records to generate

• uint32 record_size_bytes = 2

  Size of a generated record in bytes

Used in: Request

• uint64 id = 1

• uint32 level = 2

Used in: Response

• uint64 id = 1

• bytes stats_json = 2

* Similar to simple partitioning but instead of using the whole key, only the warehouse id portion of the key is used for partitioning.

Used in: Configuration

• int32 warehouses = 1

  Number of warehouses