SERVER-37853 Create sharded_core_txn with balancer suite

This reverts commit 907c8c786e26414735892bfbe763cd67e7f7e7e4.

2 files changed, 472 insertions, 0 deletions

diff --git a/buildscripts/resmokeconfig/suites/multi_stmt_txn_jscore_passthrough_with_migration.yml b/buildscripts/resmokeconfig/suites/multi_stmt_txn_jscore_passthrough_with_migration.yml
new file mode 100644
index 00000000000..8d8a2457b99
--- /dev/null
+++ b/buildscripts/resmokeconfig/suites/multi_stmt_txn_jscore_passthrough_with_migration.yml
@@ -0,0 +1,450 @@
+test_kind: multi_stmt_txn_passthrough
+
+selector:
+  roots:
+  - jstests/core/**/*.js
+  exclude_files:
+  # These tests run in the jscore_txn passthrough suites.
+  - jstests/core/txns/**/*.js
+  
+  # The following tests fail because a certain command or functionality is not supported by
+  # mongos. This command or functionality is placed in a comment next to the failing test.
+  - jstests/core/apitest_db.js  # serverStatus output doesn't have storageEngine.
+  - jstests/core/apitest_db_profile_level.js  # profiling.
+  - jstests/core/apply_ops*.js  # applyOps, SERVER-1439.
+  - jstests/core/awaitdata_getmore_cmd.js  # capped collections.
+  - jstests/core/bypass_doc_validation.js  # sharded $out output not permitted
+  - jstests/core/capped*.js  # capped collections.
+  - jstests/core/check_shard_index.js  # checkShardingIndex.
+  - jstests/core/collection_truncate.js  # emptycapped.
+  - jstests/core/collmod_without_uuid.js # applyOps, SERVER-1439
+  - jstests/core/compact_keeps_indexes.js  # compact.
+  - jstests/core/currentop.js  # uses fsync.
+  - jstests/core/dbadmin.js  # "local" database.
+  - jstests/core/dbhash.js  # dbhash.
+  - jstests/core/dbhash2.js  # dbhash.
+  - jstests/core/diagdata.js # Command not supported in mongos
+  - jstests/core/do_txn*.js  # doTxn
+  - jstests/core/dropdb_race.js  # syncdelay.
+  - jstests/core/fsync.js  # uses fsync.
+  - jstests/core/geo_haystack*.js  # geoSearch.
+  - jstests/core/geo_s2cursorlimitskip.js  # profiling.
+  - jstests/core/geo_update_btree2.js  # notablescan.
+  - jstests/core/index_id_options.js  # "local" database.
+  - jstests/core/index9.js  # "local" database.
+  - jstests/core/max_time_ms.js  # sleep, SERVER-2212.
+  - jstests/core/mr_replaceIntoDB.js  # MapReduceResult, SERVER-20495.
+  - jstests/core/notablescan.js  # notablescan.
+  - jstests/core/profile*.js  # profiling.
+  - jstests/core/queryoptimizera.js  # "local" database. 
+  - jstests/core/rename*.js # renameCollection.
+  - jstests/core/stages*.js  # stageDebug.
+  - jstests/core/startup_log.js  # "local" database.
+  - jstests/core/tailable_cursor_invalidation.js # capped collections.
+  - jstests/core/tailable_getmore_batch_size.js # capped collections.
+  - jstests/core/tailable_skip_limit.js # capped collections.
+  - jstests/core/top.js  # top.
+  # The following tests fail because mongos behaves differently from mongod when testing certain
+  # functionality. The differences are in a comment next to the failing test.
+  - jstests/core/explain_missing_database.js  # Behavior with no db different on mongos, SERVER-18047.
+  - jstests/core/geo_2d_explain.js  # executionSuccess in different spot in explain().
+  - jstests/core/geo_s2explain.js  # inputStage in different spot in explain().
+  - jstests/core/geo_s2sparse.js  # keysPerIndex in different spot in validate().
+  - jstests/core/operation_latency_histogram.js  # Stats are counted differently on mongos, SERVER-24880.
+  # Test expects to run against a standalone, but shards in this passthrough are repl sets.
+  - jstests/core/read_after_optime.js
+  # The following tests fail because they count indexes. These counts do not take into account the
+  # additional hashed shard key indexes that are automatically added by this passthrough.
+  - jstests/core/apitest_dbcollection.js
+  - jstests/core/bad_index_plugin.js
+  - jstests/core/create_indexes.js
+  - jstests/core/list_indexes_non_existent_ns.js
+  - jstests/core/rename6.js
+  # The following tests fail because they expect no databases to be created. However a DB is created
+  # automatically when we shard a collection.
+  - jstests/core/dbcase.js
+  - jstests/core/dbcase2.js
+  - jstests/core/no_db_created.js
+  - jstests/core/ns_length.js
+  - jstests/core/views/*.js # Views tests aren't expected to work when collections are implicitly sharded.
+  - jstests/core/killop_drop_collection.js # Uses fsyncLock.
+  # These tests fail because sharded clusters do not clean up correctly after failed index builds.
+  # See SERVER-33207 as an example.
+  - jstests/core/geo_borders.js
+  # TODO: Remove after fixing SERVER-32555. executionStats.nReturned is incorrect for sharded
+  # queries with a limit or for distinct commands.
+  - jstests/core/distinct_index1.js
+  - jstests/core/explain1.js
+  - jstests/core/explain4.js
+  - jstests/core/sortk.js
+  # TODO: Remove after fixing SERVER-32563. The implementation of explain for the count command is
+  # incorrect on sharded collections.
+  - jstests/core/explain_count.js
+  # TODO: Remove after fixing SERVER-14324. mapReduce behaves unpredictably when the out collection
+  # is sharded on anything other than {_id: 1}.
+  - jstests/core/mr_undef.js
+  # TODO SERVER-32311: These tests use plan stage helpers which can't handle sharded explain output.
+  - jstests/core/expr_index_use.js
+  - jstests/core/index_multikey.js
+  - jstests/core/optimized_match_explain.js
+  - jstests/core/sort_array.js
+  # TODO: SERVER-16605 MapReduce into sharded collection with hashed index fails
+  - jstests/core/mr1.js
+  - jstests/core/mr3.js
+  - jstests/core/mr4.js
+  # Create index command arbitrarily picks one of the errors if there are multiple errors from the
+  # shards. In this suite, it is possible that one of the shards don't have the collection created
+  # so it can return cannot implicitly create collection.
+  - jstests/core/index_create_too_many.js
+  - jstests/core/index_create_with_nul_in_name.js
+
+  ##
+  ## Limitations with the way the runner file injects transactions.
+  ##
+  
+  # These tests expects some statements to error, which will cause txns to abort entirely.
+  - jstests/core/bulk_api_ordered.js
+  - jstests/core/bulk_api_unordered.js
+  - jstests/core/bulk_legacy_enforce_gle.js
+  - jstests/core/capped5.js
+  - jstests/core/commands_with_uuid.js
+  - jstests/core/explain_execution_error.js
+  - jstests/core/expr.js
+  - jstests/core/find9.js
+  - jstests/core/find_getmore_bsonsize.js
+  - jstests/core/find_getmore_cmd.js
+  - jstests/core/geo_allowedcomparisons.js
+  - jstests/core/geo_big_polygon2.js
+  - jstests/core/geonear_key.js
+  - jstests/core/in.js
+  - jstests/core/index8.js # No explicit check for failed command.
+  - jstests/core/index_decimal.js
+  - jstests/core/index_multiple_compatibility.js
+  - jstests/core/index_partial_write_ops.js
+  - jstests/core/indexa.js # No explicit check for failed command.
+  - jstests/core/indexes_multiple_commands.js
+  - jstests/core/js2.js
+  - jstests/core/json_schema/json_schema.js
+  - jstests/core/mr_bigobject.js
+  - jstests/core/not2.js
+  - jstests/core/notablescan.js
+  - jstests/core/null_query_semantics.js
+  - jstests/core/or1.js
+  - jstests/core/or2.js
+  - jstests/core/or3.js
+  - jstests/core/orj.js
+  - jstests/core/ref.js
+  - jstests/core/ref4.js
+  - jstests/core/regex_limit.js
+  - jstests/core/remove_undefined.js
+  - jstests/core/set7.js 
+  - jstests/core/sortb.js
+  - jstests/core/sortf.js
+  - jstests/core/sortg.js
+  - jstests/core/sortj.js
+  - jstests/core/tailable_skip_limit.js
+  - jstests/core/type_array.js
+  - jstests/core/uniqueness.js
+  - jstests/core/unset2.js
+  - jstests/core/update_addToSet.js
+  - jstests/core/update_arrayFilters.js
+  - jstests/core/update_find_and_modify_id.js
+  - jstests/core/update_modifier_pop.js
+  - jstests/core/updateh.js
+  - jstests/core/updatej.js
+
+  # SERVER-38096 Mongos opCounter doesn't increase when last op causes the transaction to implicitly abort
+  - jstests/core/opcounters_write_cmd.js
+
+  # SERVER-36311 Remove txn_override.js restriction of snapshot readConcern
+  - jstests/core/read_after_optime.js
+
+  # Consecutive writes totalling more than the size upper threshold will cause the txn to
+  # abort with a TransactionTooLarge error.
+  - jstests/core/batch_size.js
+  - jstests/core/single_batch.js
+  # For insert1.js, only Windows DEBUG builds are affected. We compute the in-memory size
+  # of the transaction, which uses the size of the oplog entry and is 8 bytes larger on
+  # Windows debug builds than other builds.
+  - jstests/core/insert1.js
+
+  # TODO: SERVER-38207 Cannot insert document with MaxKey shard key.
+  - jstests/core/type8.js
+
+  # TODO: SERVER-38326 mapReduce cleanup doesn't obey writeConcern, rendering
+  # txn_passthrough_cmd_massage.js ineffective for preventing 2 phase drops
+  # from contesting db/coll locks.
+  - jstests/core/function_string_representations.js
+  - jstests/core/geo_mapreduce2.js
+  - jstests/core/mr*.js
+  - jstests/core/or4.js
+  - jstests/core/temp_cleanup.js
+
+  # TODO: SERVER-38570 txn_override.js does not check writeError field.
+  - jstests/core/update_setOnInsert.js
+  
+  # TODO: SERVER-38588 Hybrid index builds races with prepare in replication.
+  - jstests/core/background_index_multikey.js
+
+  # TODO: SERVER-38686 createIndex should prioritize other errors over
+  # CannotImplicitlyCreateCollection
+  - jstests/core/fts_spanish.js
+  - jstests/core/geo_invalid_polygon.js
+  - jstests/core/geo_s2index.js
+  - jstests/core/wildcard_index_validindex.js
+
+  # TODO: SERVER-38215 delx.js touches 2 different collections, which makes it susceptible for
+  # migration to happen in between and cause the transaction to abort due to StaleConfigException
+  # because it is not the first op of the transaction.
+  - jstests/core/delx.js
+
+  # benchRun bypasses the txn_override.js so it doesn't run under the same transaction and can
+  # cause test to timeout because of the 3 hr transactions lifetime setting and moveChunk
+  # getting blocked trying to acquire collection X lock while doing chunk refresh.
+  - jstests/core/bench_test*.js
+
+  ##
+  ## Some aggregation stages don't support snapshot readconcern.
+  ##
+
+  # $explain (requires read concern local)
+  - jstests/core/agg_hint.js
+  - jstests/core/and.js
+  - jstests/core/collation.js
+  - jstests/core/explain_shell_helpers.js
+  - jstests/core/index_partial_read_ops.js
+  - jstests/core/optimized_match_explain.js
+  - jstests/core/sort_array.js
+  - jstests/core/views/views_collation.js
+  - jstests/core/wildcard_index_count.js
+
+  # $out
+  - jstests/core/bypass_doc_validation.js
+  - jstests/core/views/views_aggregation.js
+
+  # $listSessions
+  - jstests/core/list_all_local_sessions.js
+  - jstests/core/list_all_sessions.js
+  - jstests/core/list_local_sessions.js
+  - jstests/core/list_sessions.js
+
+  # $indexStats
+  - jstests/core/index_stats.js
+
+  # $collStats
+  - jstests/core/operation_latency_histogram.js
+  - jstests/core/views/views_coll_stats.js
+  - jstests/core/views/views_stats.js
+
+  ##
+  ## WriteErrors get converted to WriteCommandErrors if part of a txn.
+  ##
+
+  # The same error code, but with ok:0.
+  - jstests/core/json_schema/additional_items.js
+  - jstests/core/json_schema/additional_properties.js
+  - jstests/core/json_schema/bsontype.js
+  - jstests/core/json_schema/dependencies.js
+  - jstests/core/json_schema/items.js
+  - jstests/core/json_schema/logical_keywords.js
+  - jstests/core/json_schema/min_max_items.js
+  - jstests/core/json_schema/min_max_properties.js
+  - jstests/core/json_schema/pattern_properties.js
+  - jstests/core/json_schema/required.js
+  - jstests/core/json_schema/unique_items.js
+
+  - jstests/core/field_name_validation.js
+  - jstests/core/fts_array.js
+  - jstests/core/inc-SERVER-7446.js
+  - jstests/core/invalid_db_name.js
+  - jstests/core/push_sort.js
+
+  # Checks for "WriteErrors" explicitly from the response of db.runCommand()
+  - jstests/core/max_doc_size.js
+
+  # Calls res.getWriteError() or res.hasWriteError().
+  - jstests/core/bulk_api_ordered.js
+  - jstests/core/bulk_api_unordered.js
+  - jstests/core/bulk_legacy_enforce_gle.js
+  - jstests/core/cappeda.js
+  - jstests/core/doc_validation.js
+  - jstests/core/doc_validation_options.js
+  - jstests/core/geo_multinest0.js
+  - jstests/core/insert_illegal_doc.js
+  - jstests/core/ns_length.js
+  - jstests/core/push2.js
+  - jstests/core/remove6.js
+  - jstests/core/removeb.js
+  - jstests/core/rename4.js
+  - jstests/core/shell_writeconcern.js
+  - jstests/core/storefunc.js
+  - jstests/core/update_arrayFilters.js
+  - jstests/core/update_dbref.js
+  - jstests/core/updatel.js
+  - jstests/core/write_result.js
+
+  # Multiple writes in a txn, some of which fail because the collection doesn't exist.
+  # We create the collection and retry the last write, but previous writes would have
+  # still failed.
+  - jstests/core/dbref1.js
+  - jstests/core/dbref2.js
+  - jstests/core/ref3.js
+  - jstests/core/update3.js
+  - jstests/core/rename3.js
+
+  ##
+  ## Error: Unable to acquire lock within a max lock request timeout of '0ms' milliseconds
+  ##
+
+  # Collection drops done through applyOps are not converted to w:majority
+  - jstests/core/views/invalid_system_views.js
+
+  ##
+  ## Misc. reasons.
+  ##
+
+  # SERVER-34868 Cannot run a legacy query on a session.
+  - jstests/core/exhaust.js
+  - jstests/core/validate_cmd_ns.js
+
+  # SERVER-34772 Tailable Cursors are not allowed with snapshot readconcern.
+  - jstests/core/awaitdata_getmore_cmd.js
+  - jstests/core/getmore_cmd_maxtimems.js
+  - jstests/core/tailable_cursor_invalidation.js
+  - jstests/core/tailable_getmore_batch_size.js
+
+  # SERVER-34918 The "max" option of a capped collection can be exceeded until the next insert.
+  # The reason is that we don't update the count of a collection until a transaction commits,
+  # by which point it is too late to complain that "max" has been exceeded.
+  - jstests/core/capped_max1.js
+
+  # The "max" option of a capped collection can be temporarily exceeded before a
+  # txn is committed.
+  - jstests/core/bulk_insert_capped.js
+
+  # Wrong count for top info (WriteLock)
+  - jstests/core/top.js
+
+  # Expects collection to not have been created
+  - jstests/core/insert_id_undefined.js
+
+  # Creates sessions explicitly, resulting in txns being run through different sessions
+  # using a single txnNumber.
+  - jstests/core/json_schema/misc_validation.js
+  - jstests/core/views/views_all_commands.js
+
+  # Fails with implicit sessions because it will use multiple sessions on the same Mongo connection.
+  - jstests/core/dropdb.js
+
+  # Committing a transaction when the server is fsync locked fails.
+  - jstests/core/fsync.js
+
+  # Expects legacy errors ($err).
+  - jstests/core/constructors.js
+
+  # txn interrupted by command outside of txn before getMore runs.
+  - jstests/core/commands_namespace_parsing.js
+  - jstests/core/drop3.js
+  - jstests/core/ensure_sorted.js
+  - jstests/core/geo_s2cursorlimitskip.js
+  - jstests/core/getmore_invalidated_cursors.js
+  - jstests/core/getmore_invalidated_documents.js
+  - jstests/core/kill_cursors.js
+  - jstests/core/list_collections1.js
+  - jstests/core/list_indexes.js
+  - jstests/core/max_time_ms.js
+  - jstests/core/oro.js
+
+  # Expects certain number of operations in the system.profile collection.
+  - jstests/core/profile*.js
+
+  # Parallel Shell - we do not signal the override to end a txn when a parallel shell closes.
+  - jstests/core/awaitdata_getmore_cmd.js
+  - jstests/core/compact_keeps_indexes.js
+  - jstests/core/count10.js
+  - jstests/core/count_plan_summary.js
+  - jstests/core/coveredIndex3.js
+  - jstests/core/currentop.js
+  - jstests/core/distinct3.js
+  - jstests/core/find_and_modify_concurrent_update.js
+  - jstests/core/fsync.js
+  - jstests/core/geo_update_btree.js
+  - jstests/core/killop_drop_collection.js
+  - jstests/core/loadserverscripts.js
+  - jstests/core/mr_killop.js
+  - jstests/core/queryoptimizer3.js
+  - jstests/core/remove9.js
+  - jstests/core/removeb.js
+  - jstests/core/removec.js
+  - jstests/core/shellstartparallel.js
+  - jstests/core/updatef.js
+
+  # Command expects to see result from parallel operation.
+  # E.g. Suppose the following sequence of events: op1, join() op2 in parallel shell, op3.
+  # op3 will still be using the snapshot from op1, and not see op2 at all.
+  - jstests/core/cursora.js
+  - jstests/core/bench_test1.js
+
+  exclude_with_any_tags:
+  # "Cowardly refusing to override read concern of command: ..."
+  - assumes_read_concern_unchanged
+  # "writeConcern is not allowed within a multi-statement transaction"
+  - assumes_write_concern_unchanged
+  - assumes_against_mongod_not_mongos
+  # Tests tagged with the following will fail because they assume collections are not sharded.
+  - assumes_no_implicit_collection_creation_after_drop
+  - assumes_no_implicit_index_creation
+  - assumes_unsharded_collection
+  - cannot_create_unique_index_when_using_hashed_shard_key
+  - assumes_balancer_off
+  - requires_fastcount
+
+executor:
+  config:
+    shell_options:
+      eval: >-
+        var testingReplication = true;
+        load('jstests/libs/override_methods/implicitly_shard_accessed_collections.js');
+        ImplicitlyShardAccessCollSettings.setMode(ImplicitlyShardAccessCollSettings.Modes.kHashedMoveToSingleShard);
+        load('jstests/libs/override_methods/enable_sessions.js');
+        load('jstests/libs/override_methods/txn_passthrough_cmd_massage.js');
+        load('jstests/libs/txns/txn_override.js');
+      global_vars:
+        TestData:
+          sessionOptions:
+            # Tests in this suite only read from primaries and only one node is electable, so causal
+            # consistency is not required to read your own writes.   
+            causalConsistency: false
+      readMode: commands
+  hooks:
+  # 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: CheckReplDBHash
+  - class: ValidateCollections
+  - class: DropShardedCollections
+  - class: CleanEveryN
+    n: 20
+  fixture:
+    class: ShardedClusterFixture
+    mongos_options:
+      set_parameters:
+        enableTestCommands: 1
+    mongod_options:
+      set_parameters:
+        logComponentVerbosity:
+          verbosity: 0
+          command: 2
+          replication:
+            election: 4
+            heartbeats: 2
+            initialSync: 2
+            rollback: 2
+          storage:
+            recovery: 2
+          transaction: 4
+        enableTestCommands: 1
+        numInitialSyncAttempts: 1
+        writePeriodicNoops: true
+    num_rs_nodes_per_shard: 2
+    num_shards: 2
diff --git a/buildscripts/resmokelib/testing/hooks/drop_sharded_collections.py b/buildscripts/resmokelib/testing/hooks/drop_sharded_collections.py
new file mode 100644
index 00000000000..90752571909
--- /dev/null
+++ b/buildscripts/resmokelib/testing/hooks/drop_sharded_collections.py
@@ -0,0 +1,22 @@
+"""Hook for cleaning up sharded collections created during tests."""
+
+from __future__ import absolute_import
+
+import os.path
+
+from . import jsfile
+
+
+class DropShardedCollections(jsfile.JSHook):
+    """Drops all sharded collections.
+
+    With the exception of internal collections like config.system.sessions.
+    """
+
+    def __init__(  # pylint: disable=super-init-not-called
+            self, hook_logger, fixture, shell_options=None):
+        """."""
+        description = "Drop all sharded collections"
+        js_filename = os.path.join("jstests", "hooks", "drop_sharded_collections.js")
+        jsfile.JSHook.__init__(  # pylint: disable=non-parent-init-called
+            self, hook_logger, fixture, js_filename, description, shell_options=shell_options)