test_kind: js_test
# This suite starts a 5-node replica set and runs both the DoReconfigInBackground and
# ContinuousStepdown hook with kill: true and background_reconfig: true to periodically run
# safe reconfigs and send kill signals to the primary.
# This tests that the concurrency between killing the primary and reconfigs is still
# ultimately safe.

selector:
  roots:
  - jstests/core/**/*.js
  exclude_files:
  # Transactions do not support retryability of individual operations.
  # TODO: Remove this once it is supported (SERVER-33952).
  - jstests/core/txns/**/*.js
  # The set_param1.js test attempts to compare the response from running the {getParameter: "*"}
  # command multiple times, which may observe the change to the "transactionLifetimeLimitSeconds"
  # server parameter.
  - jstests/core/set_param1.js

  # No-op retries are not ignored by top, the profiler, or opcount.
  - jstests/core/operation_latency_histogram.js
  - jstests/core/profile2.js
  - jstests/core/profile3.js
  - jstests/core/profile_findandmodify.js
  - jstests/core/top.js
  - jstests/core/views/views_stats.js

  - jstests/core/bulk_legacy_enforce_gle.js

  # TODO SERVER-31242: findAndModify no-op retry should respect the fields option.
  - jstests/core/crud_api.js
  - jstests/core/find_and_modify.js
  - jstests/core/find_and_modify2.js
  - jstests/core/find_and_modify_pipeline_update.js
  - jstests/core/find_and_modify_server6865.js

  # These test run commands using legacy queries, which are not supported on sessions.
  - jstests/core/comment_field.js
  - jstests/core/exhaust.js
  - jstests/core/validate_cmd_ns.js

  # Stepdown commands during fsync lock will fail.
  - jstests/core/currentop.js
  - jstests/core/fsync.js
  - jstests/core/killop_drop_collection.js

  # Spawns new mongo shells, which don't retry connecting on stepdown errors.
  - jstests/core/shell_connection_strings.js

  # Expect drops/creates to fail or have a certain response:
  - jstests/core/explain_upsert.js
  - jstests/core/indexes_multiple_commands.js

  # Expect certain responses, but retries of successfully completed commands may return
  # different values:
  - jstests/core/create_indexes.js
  - jstests/core/objid5.js

  # Unacknowledged writes prohibited in an explicit session.
  - jstests/core/batch_write_command_w0.js

  - jstests/core/bench_test*.js # benchRun() used for writes
  - jstests/core/benchrun_pipeline_updates.js # benchRun() used for writes
  - jstests/core/connection_string_validation.js # Does not expect a replica set connection string.
  - jstests/core/list_collections_filter.js # Temporary collections are dropped on failover.
  - jstests/core/top.js # Tests read commands (including getMore) against the secondary
  - jstests/core/drop3.js # getMore is not causally consistent if collection is dropped
  - jstests/core/validate_cmd_ns.js # Calls _exec() directly, not retryable.
  - jstests/core/list_collections_filter.js # Temporary collections are dropped on failover.
  - jstests/core/explain_large_bounds.js # Stepdown can timeout waiting for global lock.

  # Tests that fail for Causal Consistency as they have statements that do not support
  # non-local read concern.
  - jstests/core/collation.js

  # Inserts enough data that recovery takes more than 8 seconds, so we never get a working primary.
  - jstests/core/geo_s2ordering.js

  exclude_with_any_tags:
  - assumes_standalone_mongod
  ##
  # The next four tags correspond to the special errors thrown by the auto_retry_on_network_error.js
  # override when it refuses to run a certain command. Above each tag are the message(s) that cause
  # the tag to be warranted.
  ##
  # "Refusing to run a test that issues a getMore command since if a network error occurs during
  #   it then we won't know whether the cursor was advanced or not"
  - requires_getmore
  # "Refusing to run a test that issues non-retryable write operations since the test likely makes
  #   assertions on the write results and can lead to spurious failures if a network error occurs"
  - requires_non_retryable_writes
  # "Refusing to run a test that issues commands that are not blindly retryable"
  # "Refusing to run a test that issues an aggregation command with $out because it is not
  #   retryable"
  - requires_non_retryable_commands
  # "Refusing to run a test that issues commands that may return different values after a failover"
  # "Refusing to run a test that issues an aggregation command with explain because it may return
  #   incomplete results"
  # "Refusing to run a test that issues an aggregation command with
  #   $listLocalSessions because it relies on in-memory state that may not survive failovers"
  # "Refusing to run a test that issues a mapReduce command, because it calls std::terminate() if
  #   interrupted by a stepdown"
  - does_not_support_stepdowns
  ##
  # The next two tags correspond to the special errors thrown by the
  # set_read_and_write_concerns.js override when it refuses to replace the readConcern or
  # writeConcern of a particular command. Above each tag are the message(s) that cause the tag to be
  # warranted.
  ##
  # "Cowardly refusing to override read concern of command: ..."
  - assumes_read_concern_unchanged
  # "Cowardly refusing to override write concern of command: ..."
  - assumes_write_concern_unchanged
  ##
  # The next three tags corresponds to the special errors thrown by the
  # fail_unclean_shutdown_incompatible_commands.js override when it refuses to run commands that are
  # inaccurate after an unclean shutdown. Above each tag is the message that causes the tag to be
  # warranted.
  ##
  # "Cowardly fail if fastcount is run with a mongod that had an unclean shutdown: ..."
  - requires_fastcount
  # "Cowardly fail if dbStats is run with a mongod that had an unclean shutdown: ..."
  - requires_dbstats
  # "Cowardly fail if collStats is run with a mongod that had an unclean shutdown: ..."
  - requires_collstats
  # "Cowardly fail if unbounded dataSize is run with a mongod that had an unclean shutdown: ..."
  - requires_datasize
  ## The next tag corresponds to long running-operations, as they may exhaust their number
  # of retries and result in a network error being thrown.
  - operations_longer_than_stepdown_interval
  # Operations in the main test shell aren't guaranteed to be causally consistent with operations
  # performed earlier in a parallel shell if multiple nodes are electable because the latest
  # operation and cluster times aren't shared between shells.
  # "Cowardly refusing to run test with network retries enabled when it uses startParallelShell()"
  - uses_parallel_shell

executor:
  archive:
    hooks:
      - CheckReplDBHash
      - CheckReplOplogs
      - ValidateCollections
  config:
    shell_options:
      eval: >-
        testingReplication = true;
        load('jstests/libs/override_methods/network_error_and_txn_override.js');
        db = connect(TestData.connectionString);
        load("jstests/libs/override_methods/enable_sessions.js");
        load("jstests/libs/override_methods/set_read_and_write_concerns.js");
        load("jstests/libs/override_methods/fail_unclean_shutdown_incompatible_commands.js");
        load("jstests/libs/override_methods/fail_unclean_shutdown_start_parallel_shell.js");
      global_vars:
        TestData:
          alwaysInjectTransactionNumber: true
          defaultReadConcernLevel: "majority"
          logRetryAttempts: true
          networkErrorAndTxnOverrideConfig:
            retryOnNetworkErrors: true
            backgroundReconfigs: true
          overrideRetryAttempts: 3
          sessionOptions:
            writeConcern:
              w: "majority"
            readConcern:
              level: "majority"
            # Force DBClientRS to find the primary for non-write commands.
            readPreference:
              mode: "primary"
            retryWrites: true
      # We specify nodb so the shell used by each test will attempt to connect after loading the
      # retry logic in auto_retry_on_network_error.js.
      nodb: ""
      readMode: commands
  hooks:
  - class: DoReconfigInBackground
    shell_options:
      nodb: ""
  - class: ContinuousStepdown
    kill: true
    background_reconfig: true
  # The CheckReplDBHash hook waits until all operations have replicated to and have been applied
  # on the secondaries, so we run the ValidateCollections hook after it to ensure we're
  # validating the entire contents of the collection.
  - class: CheckReplOplogs
  - class: CheckReplDBHash
  - class: ValidateCollections
  - class: CleanEveryN
    n: 20
  fixture:
    class: ReplicaSetFixture
    mongod_options:
      enableMajorityReadConcern: ''
      wiredTigerEngineConfigString: "debug_mode=(table_logging=true)"
      set_parameters:
        enableTestCommands: 1
        enableElectionHandoff: 0
        logComponentVerbosity:
          replication:
            heartbeats: 2
    all_nodes_electable: true
    num_nodes: 5