Pubsub notificaiton setup for low downtime migration

README.md

@@ -2,6 +2,11 @@
 
 [![integration-tests-against-emulator](https://github.com/GoogleCloudPlatform/spanner-migration-tool/actions/workflows/integration-tests-against-emulator.yaml/badge.svg)](https://github.com/GoogleCloudPlatform/spanner-migration-tool/actions/workflows/integration-tests-against-emulator.yaml) [![code-coverage-check](https://github.com/GoogleCloudPlatform/spanner-migration-tool/actions/workflows/test-coverage.yaml/badge.svg)](https://github.com/GoogleCloudPlatform/spanner-migration-tool/actions/workflows/test-coverage.yaml) [![codecov](https://codecov.io/gh/GoogleCloudPlatform/spanner-migration-tool/graph/badge.svg?token=HY9RCUlxzm)](https://codecov.io/gh/GoogleCloudPlatform/spanner-migration-tool)
 
+
+> [!IMPORTANT]
+> We have changed architecture of the minimal downtime migration and added Pub/Sub notifications component. There are some changes in required permissions because of the new component. Please go through [Permissions page](https://googlecloudplatform.github.io/spanner-migration-tool/permissions.html) and [design page](https://googlecloudplatform.github.io/spanner-migration-tool/minimal) of the documentation.
+
+
 ## Overview
 
 Spanner migration tool is a stand-alone open source tool for Cloud Spanner evaluation and

conversion/conversion.go

@@ -298,10 +298,14 @@ func dataFromDatabase(ctx context.Context, sourceProfile profiles.SourceProfile,
 			if err != nil {
 				return nil, err
 			}
-			err = infoSchema.StartStreamingMigration(ctx, client, conv, streamInfo)
+			dfOutput, err := infoSchema.StartStreamingMigration(ctx, client, conv, streamInfo)
 			if err != nil {
 				return nil, err
 			}
+			dfJobId := dfOutput.JobID
+			gcloudCmd := dfOutput.GCloudCmd
+			streamingCfg, _ := streamInfo["streamingCfg"].(streaming.StreamingCfg)
+			streaming.StoreGeneratedResources(conv, streamingCfg, dfJobId, gcloudCmd, targetProfile.Conn.Sp.Project, "")
 			return bw, nil
 		}
 		return performSnapshotMigration(config, conv, client, infoSchema, internal.AdditionalDataAttributes{ShardId: ""}), nil
@@ -321,6 +325,7 @@ func dataFromDatabaseForDMSMigration() (*writer.BatchWriter, error) {
 func dataFromDatabaseForDataflowMigration(targetProfile profiles.TargetProfile, ctx context.Context, sourceProfile profiles.SourceProfile, conv *internal.Conv) (*writer.BatchWriter, error) {
 	updateShardsWithDataflowConfig(sourceProfile.Config.ShardConfigurationDataflow)
 	conv.Audit.StreamingStats.ShardToDataStreamNameMap = make(map[string]string)
+	conv.Audit.StreamingStats.ShardToPubsubIdMap = make(map[string]internal.PubsubCfg)
 	conv.Audit.StreamingStats.ShardToDataflowInfoMap = make(map[string]internal.ShardedDataflowJobResources)
 	tableList, err := common.GetIncludedSrcTablesFromConv(conv)
 	if err != nil {
@@ -348,13 +353,21 @@ func dataFromDatabaseForDataflowMigration(targetProfile profiles.TargetProfile,
 			return common.TaskResult[*profiles.DataShard]{Result: p, Err: err}
 		}
 		fmt.Printf("Initiating migration for shard: %v\n", p.DataShardId)
-
+		pubsubCfg, err := streaming.CreatePubsubResources(ctx, targetProfile.Conn.Sp.Project, streamingCfg.DatastreamCfg.DestinationConnectionConfig, targetProfile.Conn.Sp.Dbname)
+		if err != nil {
+			return common.TaskResult[*profiles.DataShard]{Result: p, Err: err}
+		}
+		streamingCfg.PubsubCfg = *pubsubCfg
 		err = streaming.LaunchStream(ctx, sourceProfile, p.LogicalShards, targetProfile.Conn.Sp.Project, streamingCfg.DatastreamCfg)
 		if err != nil {
 			return common.TaskResult[*profiles.DataShard]{Result: p, Err: err}
 		}
 		streamingCfg.DataflowCfg.DbNameToShardIdMap = dbNameToShardIdMap
-		err = streaming.StartDataflow(ctx, targetProfile, streamingCfg, conv)
+		dfOutput, err := streaming.StartDataflow(ctx, targetProfile, streamingCfg, conv)
+		if err != nil {
+			return common.TaskResult[*profiles.DataShard]{Result: p, Err: err}
+		}
+		streaming.StoreGeneratedResources(conv, streamingCfg, dfOutput.JobID, dfOutput.GCloudCmd, targetProfile.Conn.Sp.Project, p.DataShardId)
 		return common.TaskResult[*profiles.DataShard]{Result: p, Err: err}
 	}
 	_, err = common.RunParallelTasks(sourceProfile.Config.ShardConfigurationDataflow.DataShards, 20, asyncProcessShards, true)

docs/index.md

@@ -17,6 +17,9 @@ Spanner migration tool (SMT) is a stand-alone open source tool for Cloud Spanner
 
 ---
 
+{: .highlight }
+We have changed architecture of the minimal downtime migration and added Pub/Sub notifications component. There are changes on required permissions to run the migrations because of the new component. Please go through [Permissions page](./permissions.md) and [design page](./minimal/minimal.md) of the documentation.
+
 Spanner migration tool is a stand-alone open source tool for Cloud Spanner evaluation and
 migration, using data from an existing PostgreSQL, MySQL, SQL Server, Oracle or DynamoDB database.
 The tool ingests schema and data from either a pg_dump/mysqldump file or directly

docs/minimal/minimal.md

@@ -12,7 +12,7 @@ permalink: /minimal
 {: .note }
 Minimal downtime migrations are only supported for MySQL, Postgres and Oracle source databases.
 
-A minimal downtime migration consists of two components, migration of existing data from the database and the stream of changes (writes and updates) that are made to the source database during migration, referred to as change database capture (CDC). Using Spanner migration tool, the entire process where Datastream reads data from the source database and writes to a GCS bucket and data flow reads data from GCS bucket and writes to spanner database can be orchestrated using a unified interface. Performing schema changes on the source database during the migration is not supported. This is the suggested mode of migration for most databases.
+A minimal downtime migration consists of two components, migration of existing data from the database and the stream of changes (writes and updates) that are made to the source database during migration, referred to as change database capture (CDC). The process of migration involves Datastream reading data from the source database and writing to a GCS bucket, then GCS publishing a notification to Pub/Sub topic on each new file, then a Dataflow job when notified by the Pub/Sub subscription about the new file, reading the data from GCS bucket and writing to spanner database. With Spanner migration tool, this entire process can be orchestrated using a unified interface. Performing schema changes on the source database during the migration is not supported. This is the suggested mode of migration for most databases.
 
 ![](https://services.google.com/fh/files/helpcenter/asset-ripjb7eowf.png)
 
@@ -23,7 +23,7 @@ Sharded migrations are currently only supported for MySQL.
 
 Spanner migration tool supports sharded migrations for MySQL. Spanner migration tool does this is by multiplexing a minimal downtime migration across multiple shards. It uses the user configured schema while transforming the data read from each shard, at the time of writing to Spanner automatically. This provides an integrated experience to perform an end-to-end sharded migration. Below is the architecture of how sharded migrations work:
 
-![](https://services.google.com/fh/files/misc/smt_shard_arch.png)
+![](https://services.google.com/fh/files/misc/hb_sharded_migrations_with_pubsub_1.png)
 
 ### Terminology
 

docs/permissions.md

@@ -39,6 +39,11 @@ Ensure that Datastream and Dataflow apis are enabled on your project.
    ```sh
    gcloud services enable storage.googleapis.com
    ```
+5. Enable the Pub/Sub api by using:
+
+   ```sh
+   gcloud services enable pubsub.googleapis.com
+   ```
 
 ### Configuring connectivity for `spanner-migration-tool`
 
@@ -127,6 +132,20 @@ Grant the user **Editor role** to create buckets in the project.
 
 Enable access to Datastream, Dataflow and Spanner using [service accounts](https://cloud.google.com/compute/docs/access/create-enable-service-accounts-for-instances).
 
+### Pub/Sub
+
+Grant the user [**Pub/Sub Editor**](https://cloud.google.com/pubsub/docs/access-control#pubsub.editor) to create Pub/Sub topic and subscription for low downtime migrations.
+
+Additionally, we need to grant Pub/Sub publisher permission to GCS service agent. This will enable GCS to push a notification to a Pub/Sub topic whenever a new file is created. Refer to [this](https://cloud.google.com/storage/docs/reporting-changes#before-you-begin) page for more details.
+1. Get the GCS service agent id using the following command:
+  ```sh
+   gcloud storage service-agent --project=<PROJECT_ID>
+  ```
+2. Grant pubsub publisher role to the service agent using the following command:
+  ```sh
+   gcloud projects add-iam-policy-binding PROJECT_ID --member=serviceAccount:<GCS_SERVICE_ACCOUNT_ID> --role=roles/pubsub.publisher
+  ```
+
 ### Other Permissions
 
 In addition to these, the `DatastreamToSpanner` pipeline created by SMT requires
@@ -142,4 +161,5 @@ the following roles as well:
   - Cloud Spanner Database user
   - Cloud Spanner Restore Admin
   - Cloud Spanner Viewer
-  - Dataflow Worker
+  - Dataflow Worker
+  - Pub/Sub Subscriber

internal/convert.go

@@ -205,6 +205,20 @@ type streamingStats struct {
 	DataflowGcloudCmd        string
 	ShardToDataStreamNameMap map[string]string
 	ShardToDataflowInfoMap   map[string]ShardedDataflowJobResources
+	PubsubCfg                PubsubCfg
+	ShardToPubsubIdMap       map[string]PubsubCfg
+}
+
+type PubsubCfg struct {
+	TopicId        string
+	SubscriptionId string
+	NotificationId string
+	BucketName     string
+}
+
+type DataflowOutput struct {
+	JobID     string
+	GCloudCmd string
 }
 
 // Stores information related to rules during schema conversion

sources/common/infoschema.go

@@ -41,7 +41,7 @@ type InfoSchema interface {
 	GetIndexes(conv *internal.Conv, table SchemaAndName, colNameIdMp map[string]string) ([]schema.Index, error)
 	ProcessData(conv *internal.Conv, tableId string, srcSchema schema.Table, spCols []string, spSchema ddl.CreateTable, additionalAttributes internal.AdditionalDataAttributes) error
 	StartChangeDataCapture(ctx context.Context, conv *internal.Conv) (map[string]interface{}, error)
-	StartStreamingMigration(ctx context.Context, client *sp.Client, conv *internal.Conv, streamInfo map[string]interface{}) error
+	StartStreamingMigration(ctx context.Context, client *sp.Client, conv *internal.Conv, streamInfo map[string]interface{}) (internal.DataflowOutput, error)
 }
 
 // SchemaAndName contains the schema and name for a table

sources/dynamodb/schema.go

@@ -219,7 +219,7 @@
 // StartStreamingMigration starts the streaming migration process by creating a seperate
 // worker thread/goroutine for each table's DynamoDB Stream. It catches Ctrl+C signal if
 // customer wants to stop the process.
-func (isi InfoSchemaImpl) StartStreamingMigration(ctx context.Context, client *sp.Client, conv *internal.Conv, latestStreamArn map[string]interface{}) error {
+func (isi InfoSchemaImpl) StartStreamingMigration(ctx context.Context, client *sp.Client, conv *internal.Conv, latestStreamArn map[string]interface{}) (internal.DataflowOutput, error) {
 	fmt.Println("Processing of DynamoDB Streams started...")
 	fmt.Println("Use Ctrl+C to stop the process.")
 
@@ -243,7 +243,7 @@
 	fillConvWithStreamingStats(streamInfo, conv)
 
 	fmt.Println("DynamoDB Streams processed successfully.")
-	return nil
+	return internal.DataflowOutput{}, nil
 }
 
 func getSchemaIndexStruct(indexName string, keySchema []*dynamodb.KeySchemaElement, colNameIdMap map[string]string) schema.Index {

sources/mysql/infoschema.go

@@ -364,10 +364,19 @@
 	mp := make(map[string]interface{})
 	var (
 		tableList []string
-		err error
+		err       error
 	)
 	tableList, err = common.GetIncludedSrcTablesFromConv(conv)
-	streamingCfg, err := streaming.StartDatastream(ctx, isi.SourceProfile, isi.TargetProfile, tableList)
+	streamingCfg, err := streaming.ReadStreamingConfig(isi.SourceProfile.Conn.Mysql.StreamingConfig, isi.TargetProfile.Conn.Sp.Dbname, tableList)
+	if err != nil {
+		return nil, fmt.Errorf("error reading streaming config: %v", err)
+	}
+	pubsubCfg, err := streaming.CreatePubsubResources(ctx, isi.TargetProfile.Conn.Sp.Project, streamingCfg.DatastreamCfg.DestinationConnectionConfig, isi.SourceProfile.Conn.Mysql.Db)
+	if err != nil {
+		return nil, fmt.Errorf("error creating pubsub resources: %v", err)
+	}
+	streamingCfg.PubsubCfg = *pubsubCfg
+	streamingCfg, err = streaming.StartDatastream(ctx, streamingCfg, isi.SourceProfile, isi.TargetProfile, tableList)
 	if err != nil {
 		err = fmt.Errorf("error starting datastream: %v", err)
 		return nil, err
@@ -378,15 +387,15 @@
 
 // StartStreamingMigration is used for automatic triggering of Dataflow job when
 // performing a streaming migration.
-func (isi InfoSchemaImpl) StartStreamingMigration(ctx context.Context, client *sp.Client, conv *internal.Conv, streamingInfo map[string]interface{}) error {
+func (isi InfoSchemaImpl) StartStreamingMigration(ctx context.Context, client *sp.Client, conv *internal.Conv, streamingInfo map[string]interface{}) (internal.DataflowOutput, error) {
 	streamingCfg, _ := streamingInfo["streamingCfg"].(streaming.StreamingCfg)
 
-	err := streaming.StartDataflow(ctx, isi.TargetProfile, streamingCfg, conv)
+	dfOutput, err := streaming.StartDataflow(ctx, isi.TargetProfile, streamingCfg, conv)
 	if err != nil {
 		err = fmt.Errorf("error starting dataflow: %v", err)
-		return err
+		return internal.DataflowOutput{}, err
 	}
-	return nil
+	return dfOutput, nil
 }
 
 func toType(dataType string, columnType string, charLen sql.NullInt64, numericPrecision, numericScale sql.NullInt64) schema.Type {

sources/oracle/infoschema.go

@@ -428,10 +428,19 @@
 	mp := make(map[string]interface{})
 	var (
 		tableList []string
-		err error
+		err       error
 	)
 	tableList, err = common.GetIncludedSrcTablesFromConv(conv)
-	streamingCfg, err := streaming.StartDatastream(ctx, isi.SourceProfile, isi.TargetProfile, tableList)
+	streamingCfg, err := streaming.ReadStreamingConfig(isi.SourceProfile.Conn.Oracle.StreamingConfig, isi.TargetProfile.Conn.Sp.Dbname, tableList)
+	if err != nil {
+		return nil, fmt.Errorf("error reading streaming config: %v", err)
+	}
+	pubsubCfg, err := streaming.CreatePubsubResources(ctx, isi.TargetProfile.Conn.Sp.Project, streamingCfg.DatastreamCfg.DestinationConnectionConfig, isi.SourceProfile.Conn.Oracle.User)
+	if err != nil {
+		return nil, fmt.Errorf("error creating pubsub resources: %v", err)
+	}
+	streamingCfg.PubsubCfg = *pubsubCfg
+	streamingCfg, err = streaming.StartDatastream(ctx, streamingCfg, isi.SourceProfile, isi.TargetProfile, tableList)
 	if err != nil {
 		err = fmt.Errorf("error starting datastream: %v", err)
 		return nil, err
@@ -442,13 +451,13 @@
 
 // StartStreamingMigration is used for automatic triggering of Dataflow job when
 // performing a streaming migration.
-func (isi InfoSchemaImpl) StartStreamingMigration(ctx context.Context, client *sp.Client, conv *internal.Conv, streamingInfo map[string]interface{}) error {
+func (isi InfoSchemaImpl) StartStreamingMigration(ctx context.Context, client *sp.Client, conv *internal.Conv, streamingInfo map[string]interface{}) (internal.DataflowOutput, error) {
 	streamingCfg, _ := streamingInfo["streamingCfg"].(streaming.StreamingCfg)
-	err := streaming.StartDataflow(ctx, isi.TargetProfile, streamingCfg, conv)
+	dfOutput, err := streaming.StartDataflow(ctx, isi.TargetProfile, streamingCfg, conv)
 	if err != nil {
-		return err
+		return internal.DataflowOutput{}, err
 	}
-	return nil
+	return dfOutput, nil
 }
 
 func toType(dataType string, typecode, elementDataType sql.NullString, charLen sql.NullInt64, numericPrecision, numericScale, elementCharMaxLen, elementNumericPrecision, elementNumericScale sql.NullInt64) schema.Type {

sources/postgres/infoschema.go

@@ -63,13 +63,22 @@
 	mp := make(map[string]interface{})
 	var (
 		tableList []string
-		err error
+		err       error
 	)
 	tableList, err = common.GetIncludedSrcTablesFromConv(conv)
 	if err != nil {
 		err = fmt.Errorf("error fetching the tableList to setup datastream migration, defaulting to all tables: %v", err)
 	}
-	streamingCfg, err := streaming.StartDatastream(ctx, isi.SourceProfile, isi.TargetProfile, tableList)
+	streamingCfg, err := streaming.ReadStreamingConfig(isi.SourceProfile.Conn.Pg.StreamingConfig, isi.TargetProfile.Conn.Sp.Dbname, tableList)
+	if err != nil {
+		return nil, fmt.Errorf("error reading streaming config: %v", err)
+	}
+	pubsubCfg, err := streaming.CreatePubsubResources(ctx, isi.TargetProfile.Conn.Sp.Project, streamingCfg.DatastreamCfg.DestinationConnectionConfig, isi.TargetProfile.Conn.Sp.Dbname)
+	if err != nil {
+		return nil, fmt.Errorf("error creating pubsub resources: %v", err)
+	}
+	streamingCfg.PubsubCfg = *pubsubCfg
+	streamingCfg, err = streaming.StartDatastream(ctx, streamingCfg, isi.SourceProfile, isi.TargetProfile, tableList)
 	if err != nil {
 		err = fmt.Errorf("error starting datastream: %v", err)
 		return nil, err
@@ -78,19 +87,17 @@
 	return mp, err
 }
 
-
-
 // StartStreamingMigration is used for automatic triggering of Dataflow job when
 // performing a streaming migration.
-func (isi InfoSchemaImpl) StartStreamingMigration(ctx context.Context, client *sp.Client, conv *internal.Conv, streamingInfo map[string]interface{}) error {
+func (isi InfoSchemaImpl) StartStreamingMigration(ctx context.Context, client *sp.Client, conv *internal.Conv, streamingInfo map[string]interface{}) (internal.DataflowOutput, error) {
 	streamingCfg, _ := streamingInfo["streamingCfg"].(streaming.StreamingCfg)
 
-	err := streaming.StartDataflow(ctx, isi.TargetProfile, streamingCfg, conv)
+	dfOutput, err := streaming.StartDataflow(ctx, isi.TargetProfile, streamingCfg, conv)
 	if err != nil {
 		err = fmt.Errorf("error starting dataflow: %v", err)
-		return err
+		return internal.DataflowOutput{}, err
 	}
-	return nil
+	return dfOutput, nil
 }
 
 // GetToDdl function below implement the common.InfoSchema interface.

sources/spanner/infoschema.go

@@ -79,8 +79,8 @@
 	return nil, nil
 }
 
-func (isi InfoSchemaImpl) StartStreamingMigration(ctx context.Context, client *spanner.Client, conv *internal.Conv, streamingInfo map[string]interface{}) error {
-	return nil
+func (isi InfoSchemaImpl) StartStreamingMigration(ctx context.Context, client *spanner.Client, conv *internal.Conv, streamingInfo map[string]interface{}) (internal.DataflowOutput, error) {
+	return internal.DataflowOutput{}, nil
 }
 
 // GetTableName returns table name.

sources/sqlserver/infoschema.go

@@ -58,8 +58,8 @@
 	return nil, nil
 }
 
-func (isi InfoSchemaImpl) StartStreamingMigration(ctx context.Context, client *sp.Client, conv *internal.Conv, streamingInfo map[string]interface{}) error {
-	return nil
+func (isi InfoSchemaImpl) StartStreamingMigration(ctx context.Context, client *sp.Client, conv *internal.Conv, streamingInfo map[string]interface{}) (internal.DataflowOutput, error) {
+	return internal.DataflowOutput{}, nil
 }
 
 // GetTableName returns table name.