@powersync/service-module-mongodb-storage 0.0.0-dev-20250901073220 → 0.0.0-dev-20250903064005

package/src/storage/implementation/MongoChecksums.ts

@@ -1,3 +1,4 @@
+import * as lib_mongo from '@powersync/lib-service-mongodb';
 import {
   addPartialChecksums,
   bson,
@@ -11,27 +12,51 @@ import {
   PartialChecksumMap,
   PartialOrFullChecksum
 } from '@powersync/service-core';
-import * as lib_mongo from '@powersync/lib-service-mongodb';
-import { logger } from '@powersync/lib-services-framework';
 import { PowerSyncMongo } from './db.js';
 
+/**
+ * Checksum calculation options, primarily for tests.
+ */
+export interface MongoChecksumOptions {
+  /**
+   * How many buckets to process in a batch when calculating checksums.
+   */
+  bucketBatchLimit?: number;
+
+  /**
+   * Limit on the number of documents to calculate a checksum on at a time.
+   */
+  operationBatchLimit?: number;
+}
+
+const DEFAULT_BUCKET_BATCH_LIMIT = 200;
+const DEFAULT_OPERATION_BATCH_LIMIT = 50_000;
+
 /**
  * Checksum query implementation.
+ *
+ * General implementation flow is:
+ * 1. getChecksums() -> check cache for (partial) matches. If not found or partial match, query the remainder using computePartialChecksums().
+ * 2. computePartialChecksums() -> query bucket_state for partial matches. Query the remainder using computePartialChecksumsDirect().
+ * 3. computePartialChecksumsDirect() -> split into batches of 200 buckets at a time -> computePartialChecksumsInternal()
+ * 4. computePartialChecksumsInternal() -> aggregate over 50_000 operations in bucket_data at a time
  */
 export class MongoChecksums {
   private cache = new ChecksumCache({
     fetchChecksums: (batch) => {
-      return this.getChecksumsInternal(batch);
+      return this.computePartialChecksums(batch);
     }
   });
 
   constructor(
     private db: PowerSyncMongo,
-    private group_id: number
+    private group_id: number,
+    private options?: MongoChecksumOptions
   ) {}
 
   /**
-   * Calculate checksums, utilizing the cache.
+   * Calculate checksums, utilizing the cache for partial checkums, and querying the remainder from
+   * the database (bucket_state + bucket_data).
    */
   async getChecksums(checkpoint: InternalOpId, buckets: string[]): Promise<ChecksumMap> {
     return this.cache.getChecksumMap(checkpoint, buckets);
@@ -42,11 +67,15 @@ export class MongoChecksums {
   }
 
   /**
-   * Calculate (partial) checksums from bucket_state and the data collection.
+   * Calculate (partial) checksums from bucket_state (pre-aggregated) and bucket_data (individual operations).
+   *
+   * Results are not cached here. This method is only called by {@link ChecksumCache.getChecksumMap},
+   * which is responsible for caching its result.
    *
-   * Results are not cached.
+   * As long as data is compacted regularly, this should be fast. Large buckets without pre-compacted bucket_state
+   * can be slow.
    */
-  private async getChecksumsInternal(batch: FetchPartialBucketChecksum[]): Promise<PartialChecksumMap> {
+  private async computePartialChecksums(batch: FetchPartialBucketChecksum[]): Promise<PartialChecksumMap> {
     if (batch.length == 0) {
       return new Map();
     }
@@ -100,7 +129,7 @@ export class MongoChecksums {
       };
     });
 
-    const queriedChecksums = await this.queryPartialChecksums(mappedRequests);
+    const queriedChecksums = await this.computePartialChecksumsDirect(mappedRequests);
 
     return new Map<string, PartialOrFullChecksum>(
       batch.map((request) => {
@@ -117,61 +146,150 @@ export class MongoChecksums {
   }
 
   /**
-   * Calculate (partial) checksums from the data collection directly.
+   * Calculate (partial) checksums from the data collection directly, bypassing the cache and bucket_state.
+   *
+   * Can be used directly in cases where the cache should be bypassed, such as from a compact job.
+   *
+   * Internally, we do calculations in smaller batches of buckets as appropriate.
+   *
+   * For large buckets, this can be slow, but should not time out as the underlying queries are performed in
+   * smaller batches.
    */
-  async queryPartialChecksums(batch: FetchPartialBucketChecksum[]): Promise<PartialChecksumMap> {
-    try {
-      return await this.queryPartialChecksumsInternal(batch);
-    } catch (e) {
-      if (e.codeName == 'MaxTimeMSExpired') {
-        logger.warn(`Checksum query timed out; falling back to slower version`, e);
-        // Timeout - try the slower but more robust version
-        return await this.queryPartialChecksumsFallback(batch);
+  public async computePartialChecksumsDirect(batch: FetchPartialBucketChecksum[]): Promise<PartialChecksumMap> {
+    // Limit the number of buckets we query for at a time.
+    const bucketBatchLimit = this.options?.bucketBatchLimit ?? DEFAULT_BUCKET_BATCH_LIMIT;
+
+    if (batch.length < bucketBatchLimit) {
+      // Single batch - no need for splitting the batch and merging results
+      return await this.computePartialChecksumsInternal(batch);
+    }
+    // Split the batch and merge results
+    let results = new Map<string, PartialOrFullChecksum>();
+    for (let i = 0; i < batch.length; i += bucketBatchLimit) {
+      const bucketBatch = batch.slice(i, i + bucketBatchLimit);
+      const batchResults = await this.computePartialChecksumsInternal(bucketBatch);
+      for (let r of batchResults.values()) {
+        results.set(r.bucket, r);
       }
-      throw lib_mongo.mapQueryError(e, 'while reading checksums');
     }
+    return results;
   }
 
-  private async queryPartialChecksumsInternal(batch: FetchPartialBucketChecksum[]): Promise<PartialChecksumMap> {
-    const filters: any[] = [];
+  /**
+   * Query a batch of checksums.
+   *
+   * We limit the number of operations that the query aggregates in each sub-batch, to avoid potential query timeouts.
+   *
+   * `batch` must be limited to DEFAULT_BUCKET_BATCH_LIMIT buckets before calling this.
+   */
+  private async computePartialChecksumsInternal(batch: FetchPartialBucketChecksum[]): Promise<PartialChecksumMap> {
+    const batchLimit = this.options?.operationBatchLimit ?? DEFAULT_OPERATION_BATCH_LIMIT;
+
+    // Map requests by bucket. We adjust this as we get partial results.
+    let requests = new Map<string, FetchPartialBucketChecksum>();
     for (let request of batch) {
-      filters.push({
-        _id: {
-          $gt: {
-            g: this.group_id,
-            b: request.bucket,
-            o: request.start ?? new bson.MinKey()
-          },
-          $lte: {
-            g: this.group_id,
-            b: request.bucket,
-            o: request.end
-          }
-        }
-      });
+      requests.set(request.bucket, request);
     }
 
-    const aggregate = await this.db.bucket_data
-      .aggregate(
-        [
-          {
-            $match: {
-              $or: filters
+    const partialChecksums = new Map<string, PartialOrFullChecksum>();
+
+    while (requests.size > 0) {
+      const filters: any[] = [];
+      for (let request of requests.values()) {
+        filters.push({
+          _id: {
+            $gt: {
+              g: this.group_id,
+              b: request.bucket,
+              o: request.start ?? new bson.MinKey()
+            },
+            $lte: {
+              g: this.group_id,
+              b: request.bucket,
+              o: request.end
             }
-          },
-          CHECKSUM_QUERY_GROUP_STAGE
-        ],
-        { session: undefined, readConcern: 'snapshot', maxTimeMS: lib_mongo.MONGO_CHECKSUM_TIMEOUT_MS }
-      )
-      // Don't map the error here - we want to keep timeout errors as-is
-      .toArray();
+          }
+        });
+      }
 
-    const partialChecksums = new Map<string, PartialOrFullChecksum>(
-      aggregate.map((doc) => {
+      // Aggregate over a max of `batchLimit` operations at a time.
+      // Let's say we have 3 buckets (A, B, C), each with 10 operations, and our batch limit is 12.
+      // Then we'll do three batches:
+      // 1. Query: A[1-end], B[1-end], C[1-end]
+      //    Returns: A[1-10], B[1-2]
+      // 2. Query: B[3-end], C[1-end]
+      //    Returns: B[3-10], C[1-4]
+      // 3. Query: C[5-end]
+      //    Returns: C[5-10]
+      const aggregate = await this.db.bucket_data
+        .aggregate(
+          [
+            {
+              $match: {
+                $or: filters
+              }
+            },
+            // sort and limit _before_ grouping
+            { $sort: { _id: 1 } },
+            { $limit: batchLimit },
+            {
+              $group: {
+                _id: '$_id.b',
+                // Historically, checksum may be stored as 'int' or 'double'.
+                // More recently, this should be a 'long'.
+                // $toLong ensures that we always sum it as a long, avoiding inaccuracies in the calculations.
+                checksum_total: { $sum: { $toLong: '$checksum' } },
+                count: { $sum: 1 },
+                has_clear_op: {
+                  $max: {
+                    $cond: [{ $eq: ['$op', 'CLEAR'] }, 1, 0]
+                  }
+                },
+                last_op: { $max: '$_id.o' }
+              }
+            }
+          ],
+          { session: undefined, readConcern: 'snapshot', maxTimeMS: lib_mongo.MONGO_CHECKSUM_TIMEOUT_MS }
+        )
+        .toArray()
+        .catch((e) => {
+          throw lib_mongo.mapQueryError(e, 'while reading checksums');
+        });
+
+      let batchCount = 0;
+      let limitReached = false;
+      for (let doc of aggregate) {
         const bucket = doc._id;
-        return [bucket, checksumFromAggregate(doc)];
-      })
-    );
+        const checksum = checksumFromAggregate(doc);
+
+        const existing = partialChecksums.get(bucket);
+        if (existing != null) {
+          partialChecksums.set(bucket, addPartialChecksums(bucket, existing, checksum));
+        } else {
+          partialChecksums.set(bucket, checksum);
+        }
+
+        batchCount += doc.count;
+        if (batchCount == batchLimit) {
+          // Limit reached. Request more in the next batch.
+          // Note that this only affects the _last_ bucket in a batch.
+          limitReached = true;
+          const req = requests.get(bucket);
+          requests.set(bucket, {
+            bucket,
+            start: doc.last_op,
+            end: req!.end
+          });
+        } else {
+          // All done for this bucket
+          requests.delete(bucket);
+        }
+        batchCount++;
+      }
+      if (!limitReached) {
+        break;
+      }
+    }
 
     return new Map<string, PartialOrFullChecksum>(
       batch.map((request) => {
@@ -197,106 +315,10 @@ export class MongoChecksums {
       })
     );
   }
-
-  /**
-   * Checksums for large buckets can run over the query timeout.
-   * To avoid this, we query in batches.
-   * This version can handle larger amounts of data, but is slower, especially for many buckets.
-   */
-  async queryPartialChecksumsFallback(batch: FetchPartialBucketChecksum[]): Promise<PartialChecksumMap> {
-    const partialChecksums = new Map<string, PartialOrFullChecksum>();
-    for (let request of batch) {
-      const checksum = await this.slowChecksum(request);
-      partialChecksums.set(request.bucket, checksum);
-    }
-
-    return partialChecksums;
-  }
-
-  private async slowChecksum(request: FetchPartialBucketChecksum): Promise<PartialOrFullChecksum> {
-    const batchLimit = 50_000;
-
-    let lowerBound = 0n;
-    const bucket = request.bucket;
-
-    let runningChecksum: PartialOrFullChecksum = {
-      bucket,
-      partialCount: 0,
-      partialChecksum: 0
-    };
-    if (request.start == null) {
-      runningChecksum = {
-        bucket,
-        count: 0,
-        checksum: 0
-      };
-    }
-
-    while (true) {
-      const filter = {
-        _id: {
-          $gt: {
-            g: this.group_id,
-            b: bucket,
-            o: lowerBound
-          },
-          $lte: {
-            g: this.group_id,
-            b: bucket,
-            o: request.end
-          }
-        }
-      };
-      const docs = await this.db.bucket_data
-        .aggregate(
-          [
-            {
-              $match: filter
-            },
-            // sort and limit _before_ grouping
-            { $sort: { _id: 1 } },
-            { $limit: batchLimit },
-            CHECKSUM_QUERY_GROUP_STAGE
-          ],
-          { session: undefined, readConcern: 'snapshot', maxTimeMS: lib_mongo.MONGO_CHECKSUM_TIMEOUT_MS }
-        )
-        .toArray();
-      const doc = docs[0];
-      if (doc == null) {
-        return runningChecksum;
-      }
-      const partial = checksumFromAggregate(doc);
-      runningChecksum = addPartialChecksums(bucket, runningChecksum, partial);
-      const isFinal = doc.count != batchLimit;
-      if (isFinal) {
-        break;
-      } else {
-        lowerBound = doc.last_op;
-      }
-    }
-    return runningChecksum;
-  }
 }
 
-const CHECKSUM_QUERY_GROUP_STAGE = {
-  $group: {
-    _id: '$_id.b',
-    // Historically, checksum may be stored as 'int' or 'double'.
-    // More recently, this should be a 'long'.
-    // $toLong ensures that we always sum it as a long, avoiding inaccuracies in the calculations.
-    checksum_total: { $sum: { $toLong: '$checksum' } },
-    count: { $sum: 1 },
-    has_clear_op: {
-      $max: {
-        $cond: [{ $eq: ['$op', 'CLEAR'] }, 1, 0]
-      }
-    },
-    last_op: { $max: '$_id.o' }
-  }
-};
-
 /**
- * Convert output of CHECKSUM_QUERY_GROUP_STAGE into a checksum.
+ * Convert output of the $group stage into a checksum.
  */
 function checksumFromAggregate(doc: bson.Document): PartialOrFullChecksum {
   const partialChecksum = Number(BigInt(doc.checksum_total) & 0xffffffffn) & 0xffffffff;

package/src/storage/implementation/MongoCompactor.ts

@@ -328,15 +328,11 @@ export class MongoCompactor {
               count: 0,
               bytes: 0
             }
-          },
-          $setOnInsert: {
-            // Only set this if we're creating the document.
-            // In all other cases, the replication process will have a set a more accurate id.
-            last_op: this.maxOpId
           }
         },
-        // We generally expect this to have been created before, but do handle cases of old unchanged buckets
-        upsert: true
+        // We generally expect this to have been created before.
+        // We don't create new ones here, to avoid issues with the unique index on bucket_updates.
+        upsert: false
       }
     });
   }
@@ -483,15 +479,25 @@ export class MongoCompactor {
    * Subset of compact, only populating checksums where relevant.
    */
   async populateChecksums() {
-    let lastId: BucketStateDocument['_id'] | null = null;
+    // This is updated after each batch
+    let lowerBound: BucketStateDocument['_id'] = {
+      g: this.group_id,
+      b: new mongo.MinKey() as any
+    };
+    // This is static
+    const upperBound: BucketStateDocument['_id'] = {
+      g: this.group_id,
+      b: new mongo.MaxKey() as any
+    };
     while (!this.signal?.aborted) {
       // By filtering buckets, we effectively make this "resumeable".
-      let filter: mongo.Filter<BucketStateDocument> = {
+      const filter: mongo.Filter<BucketStateDocument> = {
+        _id: {
+          $gt: lowerBound,
+          $lt: upperBound
+        },
         compacted_state: { $exists: false }
       };
-      if (lastId) {
-        filter._id = { $gt: lastId };
-      }
 
       const bucketsWithoutChecksums = await this.db.bucket_state
         .find(filter, {
@@ -514,12 +520,12 @@ export class MongoCompactor {
 
       await this.updateChecksumsBatch(bucketsWithoutChecksums.map((b) => b._id.b));
 
-      lastId = bucketsWithoutChecksums[bucketsWithoutChecksums.length - 1]._id;
+      lowerBound = bucketsWithoutChecksums[bucketsWithoutChecksums.length - 1]._id;
     }
   }
 
   private async updateChecksumsBatch(buckets: string[]) {
-    const checksums = await this.storage.checksums.queryPartialChecksums(
+    const checksums = await this.storage.checksums.computePartialChecksumsDirect(
       buckets.map((bucket) => {
         return {
           bucket,
@@ -550,15 +556,11 @@ export class MongoCompactor {
                 checksum: BigInt(bucketChecksum.checksum),
                 bytes: null
               }
-            },
-            $setOnInsert: {
-              // Only set this if we're creating the document.
-              // In all other cases, the replication process will have a set a more accurate id.
-              last_op: this.maxOpId
             }
           },
-          // We generally expect this to have been created before, but do handle cases of old unchanged buckets
-          upsert: true
+          // We don't create new ones here - it gets tricky to get the last_op right with the unique index on:
+          // bucket_updates: {'id.g': 1, 'last_op': 1}
+          upsert: false
         }
       });
     }

package/src/storage/implementation/MongoSyncBucketStorage.ts

@@ -31,13 +31,17 @@ import { MongoBucketStorage } from '../MongoBucketStorage.js';
 import { PowerSyncMongo } from './db.js';
 import { BucketDataDocument, BucketDataKey, BucketStateDocument, SourceKey, SourceTableDocument } from './models.js';
 import { MongoBucketBatch } from './MongoBucketBatch.js';
-import { MongoChecksums } from './MongoChecksums.js';
+import { MongoChecksumOptions, MongoChecksums } from './MongoChecksums.js';
 import { MongoCompactor } from './MongoCompactor.js';
 import { MongoParameterCompactor } from './MongoParameterCompactor.js';
 import { MongoWriteCheckpointAPI } from './MongoWriteCheckpointAPI.js';
 import { idPrefixFilter, mapOpEntry, readSingleBatch, setSessionSnapshotTime } from '../../utils/util.js';
 
 
+export interface MongoSyncBucketStorageOptions {
+  checksumOptions?: MongoChecksumOptions;
+}
+
 export class MongoSyncBucketStorage
   extends BaseObserver<storage.SyncRulesBucketStorageListener>
   implements storage.SyncRulesBucketStorage
@@ -53,14 +57,15 @@ export class MongoSyncBucketStorage
     public readonly group_id: number,
     private readonly sync_rules: storage.PersistedSyncRulesContent,
     public readonly slot_name: string,
-    writeCheckpointMode: storage.WriteCheckpointMode = storage.WriteCheckpointMode.MANAGED
+    writeCheckpointMode?: storage.WriteCheckpointMode,
+    options?: MongoSyncBucketStorageOptions
   ) {
     super();
     this.db = factory.db;
-    this.checksums = new MongoChecksums(this.db, this.group_id);
+    this.checksums = new MongoChecksums(this.db, this.group_id, options?.checksumOptions);
     this.writeCheckpointAPI = new MongoWriteCheckpointAPI({
       db: this.db,
-      mode: writeCheckpointMode,
+      mode: writeCheckpointMode ?? storage.WriteCheckpointMode.MANAGED,
       sync_rules_id: group_id
     });
   }

package/src/storage/implementation/db.ts

@@ -81,7 +81,6 @@ export class PowerSyncMongo {
     await this.locks.deleteMany({});
     await this.bucket_state.deleteMany({});
     await this.custom_write_checkpoints.deleteMany({});
-    await this.connection_report_events.deleteMany({});
   }
 
   /**
@@ -145,6 +144,20 @@ export class PowerSyncMongo {
     }
     await this.db.createCollection('connection_report_events');
   }
+
+  /**
+   * Only use in migrations and tests.
+   */
+  async createBucketStateIndex() {
+    // TODO: Implement a better mechanism to use migrations in tests
+    await this.bucket_state.createIndex(
+      {
+        '_id.g': 1,
+        last_op: 1
+      },
+      { name: 'bucket_updates', unique: true }
+    );
+  }
 }
 
 export function createPowerSyncMongo(config: MongoStorageConfig, options?: lib_mongo.MongoConnectionOptions) {

package/src/storage/implementation/models.ts

@@ -98,6 +98,10 @@ export interface BucketStateDocument {
     g: number;
     b: string;
   };
+  /**
+   * Important: There is an unique index on {'_id.g': 1, last_op: 1}.
+   * That means the last_op must match an actual op in the bucket, and not the commit checkpoint.
+   */
   last_op: bigint;
   /**
    * If set, this can be treated as "cache" of a checksum at a specific point.

package/src/utils/test-utils.ts

@@ -3,10 +3,12 @@ import { PowerSyncMongo } from '../storage/implementation/db.js';
 import { TestStorageOptions } from '@powersync/service-core';
 import { MongoReportStorage } from '../storage/MongoReportStorage.js';
 import { MongoBucketStorage } from '../storage/MongoBucketStorage.js';
+import { MongoSyncBucketStorageOptions } from '../storage/implementation/MongoSyncBucketStorage.js';
 
 export type MongoTestStorageOptions = {
   url: string;
   isCI: boolean;
+  internalOptions?: MongoSyncBucketStorageOptions;
 };
 
 export function mongoTestStorageFactoryGenerator(factoryOptions: MongoTestStorageOptions) {
@@ -25,7 +27,7 @@ export function mongoTestStorageFactoryGenerator(factoryOptions: MongoTestStorag
       await db.clear();
     }
 
-    return new MongoBucketStorage(db, { slot_name_prefix: 'test_' });
+    return new MongoBucketStorage(db, { slot_name_prefix: 'test_' }, factoryOptions.internalOptions);
   };
 }
 

package/test/src/__snapshots__/storage.test.ts.snap

@@ -1,6 +1,22 @@
 // Vitest Snapshot v1, https://vitest.dev/guide/snapshot.html
 
-exports[`Mongo Sync Bucket Storage > empty storage metrics 1`] = `
+exports[`Mongo Sync Bucket Storage - Data > empty storage metrics 1`] = `
+{
+  "operations_size_bytes": 0,
+  "parameters_size_bytes": 0,
+  "replication_size_bytes": 0,
+}
+`;
+
+exports[`Mongo Sync Bucket Storage - split buckets > empty storage metrics 1`] = `
+{
+  "operations_size_bytes": 0,
+  "parameters_size_bytes": 0,
+  "replication_size_bytes": 0,
+}
+`;
+
+exports[`Mongo Sync Bucket Storage - split operations > empty storage metrics 1`] = `
 {
   "operations_size_bytes": 0,
   "parameters_size_bytes": 0,

package/test/src/connection-report-storage.test.ts

@@ -9,9 +9,9 @@ const dates = register.REPORT_TEST_DATES;
 const factory = await INITIALIZED_MONGO_REPORT_STORAGE_FACTORY();
 
 function removeVolatileFields(
-  sdks: event_types.ClientConnection[]
+  connections: event_types.ClientConnection[]
 ): Partial<event_types.ClientConnection & { _id: string }>[] {
-  return sdks.map((sdk: Partial<event_types.ClientConnection & { _id: string }>) => {
+  return connections.map((sdk: Partial<event_types.ClientConnection & { _id: string }>) => {
     const { _id, disconnected_at, connected_at, jwt_exp, ...rest } = sdk;
     return {
       ...rest
@@ -57,9 +57,9 @@ describe('Connection reporting storage', async () => {
       user_agent: userData.user_week.user_agent
     });
 
-    const sdk = await factory.db.connection_report_events.find({ user_id: userData.user_week.user_id }).toArray();
-    expect(sdk).toHaveLength(2);
-    const cleaned = removeVolatileFields(sdk);
+    const connection = await factory.db.connection_report_events.find({ user_id: userData.user_week.user_id }).toArray();
+    expect(connection).toHaveLength(2);
+    const cleaned = removeVolatileFields(connection);
     expect(cleaned).toMatchSnapshot();
   });
 
@@ -81,14 +81,14 @@ describe('Connection reporting storage', async () => {
       user_agent: userData.user_one.user_agent
     });
 
-    const sdk = await factory.db.connection_report_events
+    const connection = await factory.db.connection_report_events
       .find({ user_id: userData.user_one.user_id, client_id: userData.user_one.client_id })
       .toArray();
-    expect(sdk).toHaveLength(1);
-    expect(new Date(sdk[0].connected_at)).toEqual(newConnectAt);
-    expect(new Date(sdk[0].jwt_exp!)).toEqual(jwtExp);
-    expect(sdk[0].disconnected_at).toBeUndefined();
-    const cleaned = removeVolatileFields(sdk);
+    expect(connection).toHaveLength(1);
+    expect(new Date(connection[0].connected_at)).toEqual(newConnectAt);
+    expect(new Date(connection[0].jwt_exp!)).toEqual(jwtExp);
+    expect(connection[0].disconnected_at).toBeUndefined();
+    const cleaned = removeVolatileFields(connection);
     expect(cleaned).toMatchSnapshot();
   });
 
@@ -111,10 +111,10 @@ describe('Connection reporting storage', async () => {
       connected_at: userData.user_three.connected_at
     });
 
-    const sdk = await factory.db.connection_report_events.find({ user_id: userData.user_three.user_id }).toArray();
-    expect(sdk).toHaveLength(1);
-    expect(new Date(sdk[0].disconnected_at!)).toEqual(disconnectAt);
-    const cleaned = removeVolatileFields(sdk);
+    const connection = await factory.db.connection_report_events.find({ user_id: userData.user_three.user_id }).toArray();
+    expect(connection).toHaveLength(1);
+    expect(new Date(connection[0].disconnected_at!)).toEqual(disconnectAt);
+    const cleaned = removeVolatileFields(connection);
     expect(cleaned).toMatchSnapshot();
   });
 
@@ -124,10 +124,10 @@ describe('Connection reporting storage', async () => {
     await factory.deleteOldConnectionData({
       date: dates.weekAgo
     });
-    const sdk = await factory.getClientConnectionReports({
+    const connection = await factory.getClientConnectionReports({
       start: dates.monthAgo,
       end: dates.now
     });
-    expect(sdk).toMatchSnapshot();
+    expect(connection).toMatchSnapshot();
   });
 });