Enable WAL mode on Sqlite (#1871)

[Write-Ahead Logging](https://sqlite.org/wal.html) is both much more performant in general, and more suited to our particular access patterns.

With a simple throughput performance test, it improves performance by a factor of 5-20x depending on the sync flag.

 version                                 | throughput
-------------------------------|-------------
mode=journal sync=normal (*)| 11 htlc/s
mode=journal sync=full| 7 htlc/s
mode=wal sync=normal| 248 htlc/s
mode=wal sync=full (**)| 62 htlc/s 

(*) previous setting
(**) new setting

I went with a conservative new setting of wal+full sync, which is both 5x more performant, and more secure than what we had before.

> In WAL mode when synchronous is NORMAL (1), the WAL file is synchronized before each checkpoint and the database file is synchronized after each completed checkpoint and the WAL file header is synchronized when a WAL file begins to be reused after a checkpoint, but no sync operations occur during most transactions. With synchronous=FULL in WAL mode, an additional sync operation of the WAL file happens after each transaction commit. The extra WAL sync following each transaction help ensure that transactions are durable across a power loss. Transactions are consistent with or without the extra syncs provided by synchronous=FULL. If durability is not a concern, then synchronous=NORMAL is normally all one needs in WAL mode.

Co-authored-by: Bastien Teinturier <[email protected]>

Author: pm47

eclair-core/pom.xml

@@ -224,7 +224,7 @@
         <dependency>
             <groupId>org.xerial</groupId>
             <artifactId>sqlite-jdbc</artifactId>
-            <version>3.27.2.1</version>
+            <version>3.34.0</version>
         </dependency>
         <dependency>
             <groupId>org.postgresql</groupId>

eclair-core/src/main/scala/fr/acinq/eclair/db/Databases.scala

@@ -27,7 +27,7 @@ import grizzled.slf4j.Logging
 
 import java.io.File
 import java.nio.file._
-import java.sql.{Connection, DriverManager}
+import java.sql.Connection
 import java.util.UUID
 import scala.concurrent.duration._
 
@@ -183,28 +183,16 @@ object Databases extends Logging {
    * Given a parent folder it creates or loads all the databases from a JDBC connection
    */
   def sqlite(dbdir: File): SqliteDatabases = {
-    dbdir.mkdir()
-    var sqliteEclair: Connection = null
-    var sqliteNetwork: Connection = null
-    var sqliteAudit: Connection = null
-    try {
-      sqliteEclair = DriverManager.getConnection(s"jdbc:sqlite:${new File(dbdir, "eclair.sqlite")}")
-      sqliteNetwork = DriverManager.getConnection(s"jdbc:sqlite:${new File(dbdir, "network.sqlite")}")
-      sqliteAudit = DriverManager.getConnection(s"jdbc:sqlite:${new File(dbdir, "audit.sqlite")}")
-      SqliteUtils.obtainExclusiveLock(sqliteEclair) // there should only be one process writing to this file
-      logger.info("successful lock on eclair.sqlite")
-      SqliteDatabases(sqliteAudit, sqliteNetwork, sqliteEclair)
-    } catch {
-      case t: Throwable =>
-        logger.error("could not create connection to sqlite databases: ", t)
-        if (sqliteEclair != null) sqliteEclair.close()
-        if (sqliteNetwork != null) sqliteNetwork.close()
-        if (sqliteAudit != null) sqliteAudit.close()
-        throw t
-    }
+    dbdir.mkdirs()
+    SqliteDatabases(
+      eclairJdbc = SqliteUtils.openSqliteFile(dbdir, "eclair.sqlite", exclusiveLock = true, journalMode = "wal", syncFlag = "full"), // there should only be one process writing to this file
+      networkJdbc = SqliteUtils.openSqliteFile(dbdir, "network.sqlite", exclusiveLock = false, journalMode = "wal", syncFlag = "normal"), // we don't need strong durability guarantees on the network db
+      auditJdbc = SqliteUtils.openSqliteFile(dbdir, "audit.sqlite", exclusiveLock = false, journalMode = "wal", syncFlag = "full")
+    )
   }
 
   def postgres(dbConfig: Config, instanceId: UUID, dbdir: File, lockExceptionHandler: LockFailureHandler = LockFailureHandler.logAndStop)(implicit system: ActorSystem): PostgresDatabases = {
+    dbdir.mkdirs()
     val database = dbConfig.getString("postgres.database")
     val host = dbConfig.getString("postgres.host")
     val port = dbConfig.getInt("postgres.port")

eclair-core/src/main/scala/fr/acinq/eclair/db/sqlite/SqliteUtils.scala

@@ -17,10 +17,32 @@
 package fr.acinq.eclair.db.sqlite
 
 import fr.acinq.eclair.db.jdbc.JdbcUtils
+import grizzled.slf4j.Logging
 
-import java.sql.Connection
+import java.io.File
+import java.sql.{Connection, DriverManager}
 
-object SqliteUtils extends JdbcUtils {
+object SqliteUtils extends JdbcUtils with Logging {
+
+  def openSqliteFile(directory: File, filename: String, exclusiveLock: Boolean, journalMode: String, syncFlag: String): Connection = {
+    var sqlite: Connection = null
+    try {
+      sqlite = DriverManager.getConnection(s"jdbc:sqlite:${new File(directory, filename)}")
+      if (exclusiveLock) {
+        obtainExclusiveLock(sqlite)
+      }
+      setJournalMode(sqlite, journalMode)
+      setSynchronousFlag(sqlite, syncFlag)
+      sqlite
+    } catch {
+      case t: Throwable =>
+        logger.error("could not create connection to sqlite databases: ", t)
+        if (sqlite != null) {
+          sqlite.close()
+        }
+        throw t
+    }
+  }
 
   /**
    * Obtain an exclusive lock on a sqlite database. This is useful when we want to make sure that only one process
@@ -29,11 +51,33 @@ object SqliteUtils extends JdbcUtils {
    * The lock will be kept until the database is closed, or if the locking mode is explicitly reset.
    */
   def obtainExclusiveLock(sqlite: Connection): Unit = synchronized {
-    val statement = sqlite.createStatement()
-    statement.execute("PRAGMA locking_mode = EXCLUSIVE")
-    // we have to make a write to actually obtain the lock
-    statement.executeUpdate("CREATE TABLE IF NOT EXISTS dummy_table_for_locking (a INTEGER NOT NULL)")
-    statement.executeUpdate("INSERT INTO dummy_table_for_locking VALUES (42)")
+    using(sqlite.createStatement()) { statement =>
+      statement.execute("PRAGMA locking_mode = EXCLUSIVE")
+      // we have to make a write to actually obtain the lock
+      statement.executeUpdate("CREATE TABLE IF NOT EXISTS dummy_table_for_locking (a INTEGER NOT NULL)")
+      statement.executeUpdate("INSERT INTO dummy_table_for_locking VALUES (42)")
+    }
+  }
+
+  /**
+   * See https://www.sqlite.org/pragma.html#pragma_journal_mode
+   */
+  def setJournalMode(sqlite: Connection, mode: String): Unit = {
+    using(sqlite.createStatement()) { statement =>
+      val res = statement.executeQuery(s"PRAGMA journal_mode=$mode")
+      res.next()
+      val currentMode = res.getString(1)
+      assert(currentMode == mode, s"couldn't activate mode=$mode")
+    }
+  }
+
+  /**
+   * See https://www.sqlite.org/pragma.html#pragma_synchronous
+   */
+  def setSynchronousFlag(sqlite: Connection, flag: String): Unit = {
+    using(sqlite.createStatement()) { statement =>
+      statement.executeUpdate(s"PRAGMA synchronous=$flag")
+    }
   }
 
 }

eclair-core/src/test/scala/fr/acinq/eclair/db/ChannelsDbSpec.scala

@@ -105,7 +105,7 @@ class ChannelsDbSpec extends AnyFunSuite {
         Future(db.updateChannelMeta(channelId, ChannelEvent.EventType.PaymentSent))
       }
       val res = Future.sequence(futures)
-      Await.result(res, 60 seconds)
+      Await.result(res, 5 minutes)
     }
   }
 

eclair-core/src/test/scala/fr/acinq/eclair/integration/PerformanceIntegrationSpec.scala

@@ -0,0 +1,110 @@
+/*
+ * Copyright 2019 ACINQ SAS
+ *
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+
+package fr.acinq.eclair.integration
+
+import akka.testkit.TestProbe
+import com.typesafe.config.ConfigFactory
+import fr.acinq.bitcoin.SatoshiLong
+import fr.acinq.eclair.MilliSatoshiLong
+import fr.acinq.eclair.channel._
+import fr.acinq.eclair.payment._
+import fr.acinq.eclair.payment.receive.MultiPartHandler.ReceivePayment
+import fr.acinq.eclair.payment.send.MultiPartPaymentLifecycle.PreimageReceived
+import fr.acinq.eclair.payment.send.PaymentInitiator
+import fr.acinq.eclair.router.Router
+import org.scalatest.Ignore
+
+import java.util.UUID
+import java.util.concurrent.Executors
+import scala.concurrent.duration._
+import scala.concurrent.{Await, ExecutionContext, Future}
+import scala.jdk.CollectionConverters._
+
+/**
+ * Created by PM on 12/07/2021.
+ */
+
+@Ignore
+class PerformanceIntegrationSpec extends IntegrationSpec {
+
+  test("start eclair nodes") {
+    val commonPerfTestConfig = ConfigFactory.parseMap(Map(
+      "eclair.max-funding-satoshis" -> 100_000_000,
+      "eclair.max-accepted-htlcs" -> Channel.MAX_ACCEPTED_HTLCS,
+      "eclair.file-backup.enabled" -> false,
+    ).asJava)
+
+    instantiateEclairNode("A", ConfigFactory.parseMap(Map("eclair.node-alias" -> "A", "eclair.server.port" -> 29730).asJava).withFallback(commonPerfTestConfig).withFallback(commonFeatures).withFallback(commonConfig)) // A's channels are private
+    instantiateEclairNode("B", ConfigFactory.parseMap(Map("eclair.node-alias" -> "B", "eclair.server.port" -> 29731).asJava).withFallback(commonPerfTestConfig).withFallback(commonFeatures).withFallback(commonConfig))
+  }
+
+  test("connect nodes") {
+    // A---B
+
+    val eventListener = TestProbe()
+    nodes.values.foreach(_.system.eventStream.subscribe(eventListener.ref, classOf[ChannelStateChanged]))
+
+    connect(nodes("A"), nodes("B"), 100_000_000 sat, 0 msat)
+
+    // confirming the funding tx
+    generateBlocks(6)
+
+    within(60 seconds) {
+      eventListener.expectMsgType[ChannelStateChanged](60 seconds).currentState == NORMAL
+    }
+  }
+
+  test("wait for channels") {
+    // Channels should now be available in the router
+    val sender = TestProbe()
+    awaitCond({
+      sender.send(nodes("A").router, Router.GetRoutingState)
+      val routingState = sender.expectMsgType[Router.RoutingState]
+      routingState.channels.nonEmpty
+    }, 60 seconds)
+  }
+
+  def sendPayment()(implicit ec: ExecutionContext): Future[PaymentSent] = Future {
+    val sender = TestProbe()
+    val amountMsat = 100_000.msat
+    // first we retrieve a payment hash from B
+    sender.send(nodes("B").paymentHandler, ReceivePayment(Some(amountMsat), "1 coffee"))
+    val pr = sender.expectMsgType[PaymentRequest]
+    // then we make the actual payment
+    sender.send(nodes("A").paymentInitiator, PaymentInitiator.SendPayment(amountMsat, pr, fallbackFinalExpiryDelta = finalCltvExpiryDelta, routeParams = integrationTestRouteParams, maxAttempts = 1))
+    val paymentId = sender.expectMsgType[UUID]
+    sender.expectMsgType[PreimageReceived]
+    val ps = sender.expectMsgType[PaymentSent]
+    assert(ps.id == paymentId)
+    ps
+  }
+
+  test("send a large number of htlcs A->B") {
+    val SENDERS_COUNT = 16
+    val PAYMENTS_COUNT = 3_000
+    val ec = ExecutionContext.fromExecutor(Executors.newFixedThreadPool(SENDERS_COUNT))
+    val start = System.currentTimeMillis()
+    val futures = (0 until PAYMENTS_COUNT).map(_ => sendPayment()(ec))
+    implicit val dummyEc: ExecutionContext = ExecutionContext.Implicits.global
+    val f = Future.sequence(futures)
+    Await.result(f, 1 hour)
+    val end = System.currentTimeMillis()
+    val duration = end - start
+    println(s"$PAYMENTS_COUNT payments in ${duration}ms ${PAYMENTS_COUNT * 1000 / duration}htlc/s (senders=$SENDERS_COUNT)")
+  }
+
+}