From de524442c2ab683d31c942c6d0b3e84a6a2a3c65 Mon Sep 17 00:00:00 2001 From: Nicholas Zolnierz Date: Wed, 23 Oct 2019 19:11:39 +0000 Subject: SERVER-43303 Re introduce mapReduce fsm tests in sharded suites --- .../concurrency/fsm_workloads/map_reduce_inline.js | 18 ++-- .../concurrency/fsm_workloads/map_reduce_reduce.js | 7 +- .../fsm_workloads/map_reduce_replace_remove.js | 3 +- .../map_reduce_with_chunk_migrations.js | 106 +++++++++++++++++++++ 4 files changed, 121 insertions(+), 13 deletions(-) create mode 100644 jstests/concurrency/fsm_workloads/map_reduce_with_chunk_migrations.js (limited to 'jstests') diff --git a/jstests/concurrency/fsm_workloads/map_reduce_inline.js b/jstests/concurrency/fsm_workloads/map_reduce_inline.js index 94382b1fe70..268fc1460d7 100644 --- a/jstests/concurrency/fsm_workloads/map_reduce_inline.js +++ b/jstests/concurrency/fsm_workloads/map_reduce_inline.js @@ -61,25 +61,23 @@ var $config = (function() { var transitions = {init: {mapReduce: 1}, mapReduce: {mapReduce: 1}}; - function makeDoc(keyLimit, valueLimit) { - return { - _id: new ObjectId(), - key: Random.randInt(keyLimit), - value: Random.randInt(valueLimit) - }; - } - function setup(db, collName, cluster) { var bulk = db[collName].initializeUnorderedBulkOp(); for (var i = 0; i < this.numDocs; ++i) { // TODO: this actually does assume that there are no unique indexes - var doc = makeDoc(this.numDocs / 100, this.numDocs / 10); - bulk.insert(doc); + bulk.insert({ + _id: i, + key: Random.randInt(this.numDocs / 100), + value: Random.randInt(this.numDocs / 10) + }); } var res = bulk.execute(); assertAlways.commandWorked(res); assertAlways.eq(this.numDocs, res.nInserted); + + assert.commandWorked( + db.adminCommand({setParameter: 1, internalQueryUseAggMapReduce: true})); } return { diff --git a/jstests/concurrency/fsm_workloads/map_reduce_reduce.js b/jstests/concurrency/fsm_workloads/map_reduce_reduce.js index 86d3733df9c..99207cd5832 100644 --- a/jstests/concurrency/fsm_workloads/map_reduce_reduce.js +++ b/jstests/concurrency/fsm_workloads/map_reduce_reduce.js @@ -28,6 +28,12 @@ var $config = extendWorkload($config, function($config, $super) { } $config.states.init = function init(db, collName) { + // TODO SERVER-44150: Cannot run MR with output 'reduce' in agg since the 'whenMatched' + // pipeline will always run which can cause unexpected failures in the user-specified reduce + // function. + assert.commandWorked( + db.adminCommand({setParameter: 1, internalQueryUseAggMapReduce: false})); + $super.states.init.apply(this, arguments); this.outCollName = uniqueCollectionName(prefix, this.tid); @@ -40,7 +46,6 @@ var $config = extendWorkload($config, function($config, $super) { "output collection '" + fullName + "' should exist"); var options = {finalize: this.finalizer, out: {reduce: this.outCollName}}; - var res = db[collName].mapReduce(this.mapper, this.reducer, options); assertAlways.commandWorked(res); }; diff --git a/jstests/concurrency/fsm_workloads/map_reduce_replace_remove.js b/jstests/concurrency/fsm_workloads/map_reduce_replace_remove.js index 1bd48dce230..30225fc7c0f 100644 --- a/jstests/concurrency/fsm_workloads/map_reduce_replace_remove.js +++ b/jstests/concurrency/fsm_workloads/map_reduce_replace_remove.js @@ -21,8 +21,7 @@ load('jstests/concurrency/fsm_workloads/map_reduce_replace.js'); // for $config var $config = extendWorkload($config, function($config, $super) { $config.states.remove = function remove(db, collName) { for (var i = 0; i < 20; ++i) { - var res = db[collName].remove({value: {$gte: Random.randInt(this.numDocs / 10)}}, - {justOne: true}); + var res = db[collName].remove({_id: Random.randInt(this.numDocs)}, {justOne: true}); assertAlways.commandWorked(res); assertAlways.lte(0, res.nRemoved, tojson(res)); } diff --git a/jstests/concurrency/fsm_workloads/map_reduce_with_chunk_migrations.js b/jstests/concurrency/fsm_workloads/map_reduce_with_chunk_migrations.js new file mode 100644 index 00000000000..df4cee9c472 --- /dev/null +++ b/jstests/concurrency/fsm_workloads/map_reduce_with_chunk_migrations.js @@ -0,0 +1,106 @@ +'use strict'; + +/** + * map_reduce_with_chunk_migrations.js + * + * This tests exercises mapReduce on a collection during chunk migrations. If extending this + * workload, consider overriding the following: + * + * $config.data.collWithMigrations: collection to run chunk migrations against (default is the + * input collection of the mapReduce). + * $config.state.mapReduce: function to execute the mapReduce. + * + * @tags: [ + * requires_sharding, assumes_balancer_off, + * assumes_autosplit_off, + * requires_non_retryable_writes, + * # mapReduce does not support afterClusterTime. + * does_not_support_causal_consistency + * ] + */ +load('jstests/concurrency/fsm_libs/extend_workload.js'); // for extendWorkload +load('jstests/concurrency/fsm_workloads/sharded_moveChunk_partitioned.js'); // for $config + +var $config = extendWorkload($config, function($config, $super) { + // The base setup will insert 'partitionSize' number of documents per thread, evenly + // distributing across the chunks. Documents will only have the "_id" field. + $config.data.partitionSize = 50; + $config.threadCount = 2; + $config.iterations = 100; + $config.data.numDocs = $config.data.partitionSize * $config.threadCount; + + // By default, the collection that will be sharded with concurrent chunk migrations will be the + // one that the aggregate is run against. + $config.data.collWithMigrations = $config.collName; + $config.data.resultsCollection = "map_reduce_with_chunk_migrations_out"; + + $config.transitions = { + init: {mapReduce: 1}, + mapReduce: { + moveChunk: 0.2, + mapReduce: 0.8, + }, + moveChunk: {mapReduce: 1}, + }; + + /** + * Moves a random chunk in the target collection. + */ + $config.states.moveChunk = function moveChunk(db, collName, connCache) { + $super.states.moveChunk.apply(this, [db, this.collWithMigrations, connCache]); + }; + + /** + * Executes a mapReduce with output mode "replace". + */ + $config.states.mapReduce = function mapReduce(db, collName, connCache) { + const map = function() { + emit(this._id, 1); + }; + const reduce = function(k, values) { + return Array.sum(values); + }; + + const res = db[collName].mapReduce(map, reduce, {out: {replace: this.resultsCollection}}); + assertWhenOwnColl.commandWorked(res); + + // TODO SERVER-43290 Support for cluster stats should be able to enable this check. + // assertWhenOwnColl.eq( + // this.numDocs, res.counts.output, `Expected each _id to be output once: + // ${tojson(res)}`); + }; + + /** + * Uses the base class init() to initialize this thread for both collections. + */ + $config.states.init = function init(db, collName, connCache) { + $super.states.init.apply(this, [db, collName, connCache]); + + // Init the target collection in a similar manner, if it is different than the default + // collection. + if (collName != this.collWithMigrations) { + $super.states.init.apply(this, [db, this.collWithMigrations, connCache]); + } + }; + + /** + * Initializes the aggregate collection and the target collection for chunk migrations as + * sharded with an even distribution across each thread ID. + */ + $config.setup = function setup(db, collName, cluster) { + $super.setup.apply(this, [db, collName, cluster]); + + if (collName != this.collWithMigrations) { + // Setup the target collection in a similar manner. Note that the FSM infrastructure + // will have already enabled sharded on collName, but we need to manually do it for the + // output collection. + cluster.shardCollection(db[this.collWithMigrations], this.shardKey, false); + $super.setup.apply(this, [db, this.collWithMigrations, cluster]); + } + + assert.commandWorked( + db.adminCommand({setParameter: 1, internalQueryUseAggMapReduce: true})); + }; + + return $config; +}); -- cgit v1.2.1