keith-turner commented on code in PR #3955: URL: https://github.com/apache/accumulo/pull/3955#discussion_r1396337296
########## server/base/src/main/java/org/apache/accumulo/server/util/FindCompactionTmpFiles.java: ########## @@ -0,0 +1,215 @@ +/* + * 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 + * + * https://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.accumulo.server.util; + +import java.io.IOException; +import java.util.ArrayList; +import java.util.Arrays; +import java.util.Collection; +import java.util.HashMap; +import java.util.Iterator; +import java.util.List; +import java.util.Map; +import java.util.concurrent.ExecutionException; +import java.util.concurrent.ExecutorService; +import java.util.concurrent.Executors; +import java.util.concurrent.Future; +import java.util.concurrent.TimeUnit; + +import org.apache.accumulo.core.Constants; +import org.apache.accumulo.core.clientImpl.ClientContext; +import org.apache.accumulo.core.fate.zookeeper.ZooReader; +import org.apache.accumulo.core.trace.TraceUtil; +import org.apache.accumulo.core.util.UtilWaitThread; +import org.apache.accumulo.core.volume.Volume; +import org.apache.accumulo.server.ServerContext; +import org.apache.accumulo.server.cli.ServerUtilOpts; +import org.apache.hadoop.fs.FileStatus; +import org.apache.hadoop.fs.Path; +import org.apache.zookeeper.KeeperException; +import org.apache.zookeeper.KeeperException.NoNodeException; +import org.slf4j.Logger; +import org.slf4j.LoggerFactory; + +import com.beust.jcommander.Parameter; +import com.google.common.net.HostAndPort; + +import io.opentelemetry.api.trace.Span; +import io.opentelemetry.context.Scope; + +public class FindCompactionTmpFiles { + + private static final Logger LOG = LoggerFactory.getLogger(FindCompactionTmpFiles.class); + + static class Opts extends ServerUtilOpts { + @Parameter(names = "--delete") + boolean delete = false; + } + + private static boolean allCompactorsDown(ClientContext context) { + // This is a copy of ExternalCompactionUtil.getCompactorAddrs that returns + // false if any compactor address is found. If there are no compactor addresses + // in any of the groups, then it returns true. + try { + final Map<String,List<HostAndPort>> groupsAndAddresses = new HashMap<>(); + final String compactorGroupsPath = context.getZooKeeperRoot() + Constants.ZCOMPACTORS; + ZooReader zooReader = context.getZooReader(); + List<String> groups = zooReader.getChildren(compactorGroupsPath); + for (String group : groups) { + groupsAndAddresses.putIfAbsent(group, new ArrayList<>()); + try { + List<String> compactors = zooReader.getChildren(compactorGroupsPath + "/" + group); + for (String compactor : compactors) { + // compactor is the address, we are checking to see if there is a child node which + // represents the compactor's lock as a check that it's alive. + List<String> children = + zooReader.getChildren(compactorGroupsPath + "/" + group + "/" + compactor); + if (!children.isEmpty()) { + LOG.trace("Found live compactor {} ", compactor); + return false; + } + } + } catch (NoNodeException e) { + LOG.trace("Ignoring node that went missing", e); + } + } + return true; + } catch (KeeperException e) { + throw new IllegalStateException(e); + } catch (InterruptedException e) { + Thread.currentThread().interrupt(); + throw new IllegalStateException(e); + } + } + + public static List<Path> findTempFiles(ServerContext context) throws InterruptedException { + final String pattern = "/tables/*/*/*"; + final Collection<Volume> vols = context.getVolumeManager().getVolumes(); + final ExecutorService svc = Executors.newFixedThreadPool(vols.size()); + final List<Path> matches = new ArrayList<>(1024); + final List<Future<Void>> futures = new ArrayList<>(vols.size()); + for (Volume vol : vols) { + final Path volPattern = new Path(vol.getBasePath() + pattern); + LOG.info("Looking for tmp files in volume: {} that match pattern: {}", vol, volPattern); + futures.add(svc.submit(() -> { + try { + FileStatus[] files = + vol.getFileSystem().globStatus(volPattern, (p) -> p.getName().contains("_tmp_ECID-")); + System.out.println(Arrays.toString(files)); + Arrays.stream(files).forEach(fs -> matches.add(fs.getPath())); + } catch (IOException e) { + LOG.error("Error looking for tmp files in volume: {}", vol, e); + } + return null; + })); + } + svc.shutdown(); + + while (futures.size() > 0) { + UtilWaitThread.sleep(10_000); + Iterator<Future<Void>> iter = futures.iterator(); + while (iter.hasNext()) { + Future<Void> future = iter.next(); + if (future.isDone()) { + iter.remove(); + try { + future.get(); + } catch (InterruptedException | ExecutionException e) { + throw new RuntimeException("Error getting list of tmp files", e); + } + } + } + } + svc.awaitTermination(10, TimeUnit.MINUTES); + return matches; + } + + public static class DeleteStats { + public int success = 0; + public int failure = 0; + public int error = 0; + } + + public static DeleteStats deleteTempFiles(ServerContext context, List<Path> filesToDelete) + throws InterruptedException { + + final ExecutorService delSvc = Executors.newFixedThreadPool(8); + final List<Future<Boolean>> futures = new ArrayList<>(filesToDelete.size()); + final DeleteStats stats = new DeleteStats(); + + filesToDelete.forEach(p -> { + futures.add(delSvc.submit(() -> context.getVolumeManager().delete(p))); + }); + delSvc.shutdown(); + + int expectedResponses = filesToDelete.size(); + while (expectedResponses > 0) { + UtilWaitThread.sleep(10_000); + Iterator<Future<Boolean>> iter = futures.iterator(); + while (iter.hasNext()) { + Future<Boolean> future = iter.next(); + if (future.isDone()) { + expectedResponses--; + iter.remove(); + try { + if (future.get()) { + stats.success++; + } else { + stats.failure++; + } + } catch (ExecutionException e) { + stats.error++; + LOG.error("Error deleting a compaction tmp file", e); + } + } + } + LOG.debug("Waiting on {} responses", expectedResponses); + } + delSvc.awaitTermination(10, TimeUnit.MINUTES); + return stats; + } + + public static void main(String[] args) throws Exception { + Opts opts = new Opts(); + opts.parseArgs(FindCompactionTmpFiles.class.getName(), args); + LOG.info("Deleting compaction tmp files: {}", opts.delete); + Span span = TraceUtil.startSpan(FindCompactionTmpFiles.class, "main"); + try (Scope scope = span.makeCurrent()) { + ServerContext context = opts.getServerContext(); + if (!allCompactorsDown(context)) { + LOG.warn("Compactor addresses found in ZooKeeper. Unable to run this utility."); + } + + final List<Path> matches = findTempFiles(context); + LOG.info("Found the following compaction tmp files:"); + matches.forEach(p -> LOG.info("{}", p)); + Review Comment: At this point we could scan the metadata table and get active ECIDs and remove paths from matches that have an active ECID. Scanning for only tablets that have an ECID would be another good use case for #3933. Would need to scan all three levels of metadata. That would make this utility able to be run while compactors are running. If that is the case then the compaction coordinator could possibly call this code periodically (like once a day by default). The other code for deleting temp files could miss files if the coordinator dies a certain time or because of race conditions with split and merge that delete the tablet and create new tablet. Periodically running this would still clean up things missed by the other code. -- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. To unsubscribe, e-mail: [email protected] For queries about this service, please contact Infrastructure at: [email protected]
