org.apache.accumulo.manager.tableOps.merge.TableRangeOpWait Maven / Gradle / Ivy

Go to download

Show more of this group Show more artifacts with this name
Show all versions of accumulo-manager Show documentation

The manager server for Apache Accumulo for load balancing and other system-wide operations.

There is a newer version: 3.0.0

/*
 * 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.manager.tableOps.merge;

import org.apache.accumulo.core.data.NamespaceId;
import org.apache.accumulo.core.data.TableId;
import org.apache.accumulo.core.fate.Repo;
import org.apache.accumulo.manager.Manager;
import org.apache.accumulo.manager.tableOps.ManagerRepo;
import org.apache.accumulo.manager.tableOps.Utils;
import org.apache.accumulo.server.manager.state.MergeInfo;
import org.apache.accumulo.server.manager.state.MergeState;
import org.slf4j.Logger;
import org.slf4j.LoggerFactory;

/**
 * Merge makes things hard.
 *
 * Typically, a client will read the list of tablets, and begin an operation on that tablet at the
 * location listed in the metadata table. When a tablet splits, the information read from the
 * metadata table doesn't match reality, so the operation fails, and must be retried. But the
 * operation will take place either on the parent, or at a later time on the children. It won't take
 * place on just half of the tablet.
 *
 * However, when a merge occurs, the operation may have succeeded on one section of the merged area,
 * and not on the others, when the merge occurs. There is no way to retry the request at a later
 * time on an unmodified tablet.
 *
 * The code below uses read-write lock to prevent some operations while a merge is taking place.
 * Normal operations, like bulk imports, will grab the read lock and prevent merges (writes) while
 * they run. Merge operations will lock out some operations while they run.
 */
class TableRangeOpWait extends ManagerRepo {
  private static final Logger log = LoggerFactory.getLogger(TableRangeOpWait.class);

  private static final long serialVersionUID = 1L;
  private TableId tableId;
  private NamespaceId namespaceId;

  public TableRangeOpWait(NamespaceId namespaceId, TableId tableId) {
    this.tableId = tableId;
    this.namespaceId = namespaceId;
  }

  @Override
  public long isReady(long tid, Manager env) {
    if (!env.getMergeInfo(tableId).getState().equals(MergeState.NONE)) {
      return 50;
    }
    return 0;
  }

  @Override
  public Repo call(long tid, Manager manager) throws Exception {
    MergeInfo mergeInfo = manager.getMergeInfo(tableId);
    log.info("removing merge information " + mergeInfo);
    manager.clearMergeState(tableId);
    Utils.unreserveTable(manager, tableId, tid, true);
    Utils.unreserveNamespace(manager, namespaceId, tid, false);
    return null;
  }

}