Github user zsxwing commented on a diff in the pull request: https://github.com/apache/spark/pull/13513#discussion_r78630727 --- Diff: sql/core/src/main/scala/org/apache/spark/sql/execution/streaming/CompactibleFileStreamLog.scala --- @@ -0,0 +1,249 @@ +/* + * Licensed to the Apache Software Foundation (ASF) under one or more + * contributor license agreements. See the NOTICE file distributed with + * this work for additional information regarding copyright ownership. + * The ASF licenses this file to You 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 org.apache.spark.sql.execution.streaming + +import java.io.IOException +import java.nio.charset.StandardCharsets.UTF_8 +import java.util.concurrent.TimeUnit + +import scala.reflect.ClassTag + +import org.apache.hadoop.fs.{Path, PathFilter} + +import org.apache.spark.sql.SparkSession + +/** + * An abstract class for compactible metadata logs. It will write one log file for each batch. + * The first line of the log file is the version number, and there are multiple JSON lines + * following. + * + * As reading from many small files is usually pretty slow, also too many + * small files in one folder will mess the FS, [[CompactibleFileStreamLog]] will + * compact log files every 10 batches by default into a big file. When + * doing a compaction, it will read all old log files and merge them with the new batch. + */ +abstract class CompactibleFileStreamLog[T: ClassTag]( + sparkSession: SparkSession, + path: String) + extends HDFSMetadataLog[Array[T]](sparkSession, path) { + + import CompactibleFileStreamLog._ + + /** + * If we delete the old files after compaction at once, there is a race condition in S3: other + * processes may see the old files are deleted but still cannot see the compaction file using + * "list". The `allFiles` handles this by looking for the next compaction file directly, however, + * a live lock may happen if the compaction happens too frequently: one processing keeps deleting + * old files while another one keeps retrying. Setting a reasonable cleanup delay could avoid it. + */ + protected val fileCleanupDelayMs = TimeUnit.MINUTES.toMillis(10) + + protected val isDeletingExpiredLog = true + + protected val compactInterval = 10 + + /** + * Serialize the data into encoded string. + */ + protected def serializeData(t: T): String + + /** + * Deserialize the string into data object. + */ + protected def deserializeData(encodedString: String): T + + /** + * Filter out the unwanted logs, by default it filters out nothing, inherited class could + * override this method to do filtering. + */ + protected def compactLogs(oldLogs: Seq[T], newLogs: Seq[T]): Seq[T] = { + oldLogs ++ newLogs + } + + override def batchIdToPath(batchId: Long): Path = { + if (isCompactionBatch(batchId, compactInterval)) { + new Path(metadataPath, s"$batchId$COMPACT_FILE_SUFFIX") + } else { + new Path(metadataPath, batchId.toString) + } + } + + override def pathToBatchId(path: Path): Long = { + getBatchIdFromFileName(path.getName) + } + + override def isBatchFile(path: Path): Boolean = { + try { + getBatchIdFromFileName(path.getName) + true + } catch { + case _: NumberFormatException => false + } + } + + override def serialize(logData: Array[T]): Array[Byte] = { + (VERSION +: logData.map(serializeData)).mkString("\n").getBytes(UTF_8) --- End diff -- Could you make VERSION be a constructor parameter in order to support to change source or sink format separately?
--- If your project is set up for it, you can reply to this email and have your reply appear on GitHub as well. If your project does not have this feature enabled and wishes so, or if the feature is enabled but not working, please contact infrastructure at infrastruct...@apache.org or file a JIRA ticket with INFRA. --- --------------------------------------------------------------------- To unsubscribe, e-mail: reviews-unsubscr...@spark.apache.org For additional commands, e-mail: reviews-h...@spark.apache.org