PERF-2360: Add new workload for untargeted sharded $lookup (mongodb#491)

src/workloads/execution/ShardedLookup.yml

@@ -0,0 +1,118 @@
+SchemaVersion: 2018-07-01
+Owner: "@mongodb/query"
+Description: |
+  This test exercises the behavior of $lookup against a sharded foreign collection.
+
+  The workload consists of the following phases:
+    1. Creating empty sharded collections distributed across all shards in the cluster.
+    2. Populating collections with data.
+    3. Fsync.
+    4. Running untargeted $lookup from a sharded collection.
+
+Actors:
+- Name: CreateShardedCollections
+  Type: AdminCommand
+  Threads: 1
+  Phases:
+  - Repeat: 1
+    Database: admin
+    Operations:
+    # Shard Collection0 and Collection1 using hashed sharding to ensure that the chunks are evenly
+    # distributed across the shards.
+    - OperationMetricsName: EnableSharding
+      OperationName: AdminCommand
+      OperationCommand:
+        enableSharding: &Database test
+    - OperationMetricsName: ShardLocalCollection
+      OperationName: AdminCommand
+      OperationCommand:
+        shardCollection: test.Collection0
+        key: {shardKey: hashed}
+        numInitialChunks: &NumChunks 6
+    - OperationMetricsName: ShardForeignCollection
+      OperationName: AdminCommand
+      OperationCommand:
+        shardCollection: test.Collection1
+        key: {shardKey: hashed}
+        numInitialChunks: *NumChunks
+    # Disable the balancer so that it can't skew results while the $lookups are running.
+    - OperationMetricsName: DisableBalancer
+      OperationName: AdminCommand
+      OperationCommand:
+        balancerStop: 1
+  - &Nop {Nop: true}
+  - *Nop
+  - *Nop
+
+- Name: LoadInitialData
+  Type: Loader
+  Threads: 1
+  Phases:
+  - *Nop
+  - Repeat: 1
+    BatchSize: 1000
+    Threads: 1
+    DocumentCount: &NumDocs 3000
+    Database: *Database
+    CollectionCount: 2    # Loader will populate 'Collection0' then 'Collection1'.
+    Document:
+      shardKey: {^RandomInt: {min: 1, max: 100}}
+      date: &Date {^RandomDate: {min: "2020-01-01", max: "2021-01-01"}}
+      int: {^RandomInt: {min: 1, max: 100}}
+  - *Nop
+  - *Nop
+
+- Name: Quiesce
+  Type: RunCommand
+  Threads: 1
+  Phases:
+  - *Nop
+  - *Nop
+  - Repeat: 1
+    Database: admin
+    Operations:
+    - OperationName: RunCommand
+      OperationCommand:
+        fsync: 1
+  - *Nop
+
+- Name: RunLookups
+  Type: RunCommand
+  Database: *Database
+  Phases:
+  - *Nop
+  - *Nop
+  - *Nop
+  - Repeat: 10    # Untargeted $lookup from sharded collection to sharded collection
+    Database: *Database
+    Operations:
+    - OperationMetricsName: UntargetedLookupShardedToSharded
+      OperationName: RunCommand
+      OperationCommand:
+        aggregate: Collection0
+        pipeline:
+          [{
+            $lookup: {
+              from: Collection1,
+              let: {localInt: "$int"},
+              pipeline: [{
+                $match: {
+                  $expr: {
+                    $and: [
+                      {$eq: ["$int", "$$localInt"]},
+                      {$lte: ["$date", *Date]}
+                    ]
+                  }
+                }
+              }],
+              as: matches
+            }
+          }]
+        # To get meaningful results, the entire result set should fit in a single batch. This should
+        # be possible since both collections are small.
+        cursor: {batchSize: *NumDocs}
+
+AutoRun:
+  - When:
+      mongodb_setup:
+        $eq: shard-lite-all-feature-flags