org.apache.cassandra.service.RowDataResolver Maven / Gradle / Ivy
Go to download
Show more of this group Show more artifacts with this name
Show all versions of cassandra-all Show documentation
Show all versions of cassandra-all Show documentation
Palantir open source project
/*
* 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.cassandra.service;
import java.net.InetAddress;
import java.nio.ByteBuffer;
import java.util.ArrayList;
import java.util.Collections;
import java.util.List;
import java.util.concurrent.TimeUnit;
import com.google.common.collect.Iterables;
import org.apache.cassandra.db.*;
import org.apache.cassandra.db.columniterator.IdentityQueryFilter;
import org.apache.cassandra.db.filter.IDiskAtomFilter;
import org.apache.cassandra.db.filter.QueryFilter;
import org.apache.cassandra.net.*;
import org.apache.cassandra.tracing.Tracing;
import org.apache.cassandra.utils.CloseableIterator;
import org.apache.cassandra.utils.FBUtilities;
public class RowDataResolver extends AbstractRowResolver
{
private int maxLiveCount = 0;
public List repairResults = Collections.emptyList();
private final IDiskAtomFilter filter;
private final long timestamp;
public RowDataResolver(String keyspaceName, ByteBuffer key, IDiskAtomFilter qFilter, long timestamp, int maxResponseCount)
{
super(key, keyspaceName, maxResponseCount);
this.filter = qFilter;
this.timestamp = timestamp;
}
/*
* This method handles the following scenario:
*
* there was a mismatch on the initial read, so we redid the digest requests
* as full data reads. In this case we need to compute the most recent version
* of each column, and send diffs to out-of-date replicas.
*/
public Row resolve() throws DigestMismatchException
{
int replyCount = replies.size();
if (logger.isTraceEnabled())
logger.trace("resolving {} responses", replyCount);
long start = System.nanoTime();
ColumnFamily resolved;
if (replyCount > 1)
{
List versions = new ArrayList<>(replyCount);
List endpoints = new ArrayList<>(replyCount);
for (MessageIn message : replies)
{
ReadResponse response = message.payload;
ColumnFamily cf = response.row().cf;
assert !response.isDigestQuery() : "Received digest response to repair read from " + message.from;
versions.add(cf);
endpoints.add(message.from);
// compute maxLiveCount to prevent short reads -- see https://issues.apache.org/jira/browse/CASSANDRA-2643
int liveCount = cf == null ? 0 : filter.getLiveCount(cf, timestamp);
if (liveCount > maxLiveCount)
maxLiveCount = liveCount;
}
resolved = resolveSuperset(versions, timestamp);
if (logger.isTraceEnabled())
logger.trace("versions merged");
// send updates to any replica that was missing part of the full row
// (resolved can be null even if versions doesn't have all nulls because of the call to removeDeleted in resolveSuperSet)
if (resolved != null)
repairResults = scheduleRepairs(resolved, keyspaceName, key, versions, endpoints);
}
else
{
resolved = replies.get(0).payload.row().cf;
}
if (logger.isTraceEnabled())
logger.trace("resolve: {} ms.", TimeUnit.NANOSECONDS.toMillis(System.nanoTime() - start));
return new Row(key, resolved);
}
/**
* For each row version, compare with resolved (the superset of all row versions);
* if it is missing anything, send a mutation to the endpoint it come from.
*/
public static List scheduleRepairs(ColumnFamily resolved, String keyspaceName, DecoratedKey key, List versions, List endpoints)
{
List results = new ArrayList(versions.size());
for (int i = 0; i < versions.size(); i++)
{
ColumnFamily diffCf = ColumnFamily.diff(versions.get(i), resolved);
if (diffCf == null) // no repair needs to happen
continue;
// create and send the mutation message based on the diff
Mutation mutation = new Mutation(keyspaceName, key.getKey(), diffCf);
// use a separate verb here because we don't want these to be get the white glove hint-
// on-timeout behavior that a "real" mutation gets
Tracing.trace("Sending read-repair-mutation to {}", endpoints.get(i));
results.add(MessagingService.instance().sendRR(mutation.createMessage(MessagingService.Verb.READ_REPAIR),
endpoints.get(i)));
}
return results;
}
static ColumnFamily resolveSuperset(Iterable versions, long now)
{
assert Iterables.size(versions) > 0;
ColumnFamily resolved = null;
for (ColumnFamily cf : versions)
{
if (cf == null)
continue;
if (resolved == null)
resolved = cf.cloneMeShallow();
else
resolved.delete(cf);
}
if (resolved == null)
return null;
// mimic the collectCollatedColumn + removeDeleted path that getColumnFamily takes.
// this will handle removing columns and subcolumns that are suppressed by a row or
// supercolumn tombstone.
QueryFilter filter = new QueryFilter(null, resolved.metadata().cfName, new IdentityQueryFilter(), now);
List> iters = new ArrayList<>(Iterables.size(versions));
for (ColumnFamily version : versions)
if (version != null)
iters.add(FBUtilities.closeableIterator(version.iterator()));
filter.collateColumns(resolved, iters, Integer.MIN_VALUE);
return ColumnFamilyStore.removeDeleted(resolved, Integer.MIN_VALUE);
}
public Row getData()
{
assert !replies.isEmpty();
return replies.get(0).payload.row();
}
public boolean isDataPresent()
{
return !replies.isEmpty();
}
public int getMaxLiveCount()
{
return maxLiveCount;
}
}