Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

CURATOR-688. SharedCount will be never updated successful when version of ZNode is overflow. #478

Merged
merged 4 commits into from
Jun 4, 2024
Merged
Show file tree
Hide file tree
Changes from all commits
Commits
File filter

Filter by extension

Filter by extension

Conversations
Failed to load comments.
Jump to
Jump to file
Failed to load files.
Diff view
Diff view
Original file line number Diff line number Diff line change
@@ -0,0 +1,39 @@
/*
* 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.curator.framework.recipes.shared;

import org.apache.zookeeper.data.Stat;

/**
* Exception to alert overflowed {@link Stat#getVersion()} {@code -1} which is not suitable in
* {@link SharedValue#trySetValue(VersionedValue, byte[])} and {@link SharedCount#trySetCount(VersionedValue, int)}.
*
* <p>In case of this exception, clients have to choose:
* <ul>
* <li>Take their own risk to do a blind set.</li>
* <li>Update ZooKeeper cluster to solve <a href="https://issues.apache.org/jira/browse/ZOOKEEPER-4743">ZOOKEEPER-4743</a>.</li>
* </ul>
*/
public class IllegalTrySetVersionException extends IllegalArgumentException {
@Override
public String getMessage() {
return "overflowed Stat.version -1 is not suitable for trySet(a.k.a. compare-and-set ZooKeeper::setData)";
}
}
Original file line number Diff line number Diff line change
Expand Up @@ -30,6 +30,7 @@
import org.apache.curator.framework.CuratorFramework;
import org.apache.curator.framework.listen.Listenable;
import org.apache.curator.framework.state.ConnectionState;
import org.apache.zookeeper.data.Stat;

/**
* Manages a shared integer. All clients watching the same path will have the up-to-date
Expand Down Expand Up @@ -60,7 +61,7 @@ public int getCount() {
@Override
public VersionedValue<Integer> getVersionedValue() {
VersionedValue<byte[]> localValue = sharedValue.getVersionedValue();
return new VersionedValue<Integer>(localValue.getVersion(), fromBytes(localValue.getValue()));
return localValue.mapValue(SharedCount::fromBytes);
}

/**
Expand Down Expand Up @@ -102,11 +103,11 @@ public boolean trySetCount(int newCount) throws Exception {
* @param newCount the new value to attempt
* @return true if the change attempt was successful, false if not. If the change
* was not successful, {@link #getCount()} will return the updated value
* @throws IllegalTrySetVersionException if {@link Stat#getVersion()} overflowed to {@code -1}
* @throws Exception ZK errors, interruptions, etc.
*/
public boolean trySetCount(VersionedValue<Integer> previous, int newCount) throws Exception {
VersionedValue<byte[]> previousCopy =
new VersionedValue<byte[]>(previous.getVersion(), toBytes(previous.getValue()));
VersionedValue<byte[]> previousCopy = previous.mapValue(SharedCount::toBytes);
return sharedValue.trySetValue(previousCopy, toBytes(newCount));
}

Expand Down
Original file line number Diff line number Diff line change
Expand Up @@ -48,6 +48,7 @@
* value (considering ZK's normal consistency guarantees).
*/
public class SharedValue implements Closeable, SharedValueReader {
private static final int NO_ZXID = -1;
private static final int UNINITIALIZED_VERSION = -1;

private final Logger log = LoggerFactory.getLogger(getClass());
Expand Down Expand Up @@ -101,8 +102,8 @@ public SharedValue(CuratorFramework client, String path, byte[] seedValue) {
this.path = PathUtils.validatePath(path);
this.seedValue = Arrays.copyOf(seedValue, seedValue.length);
this.watcher = new SharedValueCuratorWatcher();
currentValue = new AtomicReference<VersionedValue<byte[]>>(
new VersionedValue<byte[]>(UNINITIALIZED_VERSION, Arrays.copyOf(seedValue, seedValue.length)));
currentValue = new AtomicReference<>(
new VersionedValue<>(NO_ZXID, UNINITIALIZED_VERSION, Arrays.copyOf(seedValue, seedValue.length)));
}

@VisibleForTesting
Expand All @@ -112,8 +113,8 @@ protected SharedValue(WatcherRemoveCuratorFramework client, String path, byte[]
this.seedValue = Arrays.copyOf(seedValue, seedValue.length);
// inject watcher for testing
this.watcher = watcher;
currentValue = new AtomicReference<VersionedValue<byte[]>>(
new VersionedValue<byte[]>(UNINITIALIZED_VERSION, Arrays.copyOf(seedValue, seedValue.length)));
currentValue = new AtomicReference<>(
new VersionedValue<>(NO_ZXID, UNINITIALIZED_VERSION, Arrays.copyOf(seedValue, seedValue.length)));
}

@Override
Expand All @@ -125,12 +126,11 @@ public byte[] getValue() {
@Override
public VersionedValue<byte[]> getVersionedValue() {
VersionedValue<byte[]> localCopy = currentValue.get();
return new VersionedValue<byte[]>(
localCopy.getVersion(), Arrays.copyOf(localCopy.getValue(), localCopy.getValue().length));
return localCopy.mapValue(bytes -> Arrays.copyOf(bytes, bytes.length));
}

/**
* Change the shared value value irrespective of its previous state
* Change the shared value irrespective of its previous state
*
* @param newValue new value
* @throws Exception ZK errors, interruptions, etc.
Expand All @@ -139,7 +139,7 @@ public void setValue(byte[] newValue) throws Exception {
Preconditions.checkState(state.get() == State.STARTED, "not started");

Stat result = client.setData().forPath(path, newValue);
updateValue(result.getVersion(), Arrays.copyOf(newValue, newValue.length));
updateValue(result.getMzxid(), result.getVersion(), Arrays.copyOf(newValue, newValue.length));
}

/**
Expand Down Expand Up @@ -171,19 +171,25 @@ public boolean trySetValue(byte[] newValue) throws Exception {
* @param newValue the new value to attempt
* @return true if the change attempt was successful, false if not. If the change
* was not successful, {@link #getValue()} will return the updated value
* @throws IllegalTrySetVersionException if {@link Stat#getVersion()} overflowed to {@code -1}
* @throws Exception ZK errors, interruptions, etc.
*/
public boolean trySetValue(VersionedValue<byte[]> previous, byte[] newValue) throws Exception {
Preconditions.checkState(state.get() == State.STARTED, "not started");

VersionedValue<byte[]> current = currentValue.get();
if (previous.getVersion() != current.getVersion() || !Arrays.equals(previous.getValue(), current.getValue())) {
// Omit comparing of getVersion here, so we can test the exception case.
// This affects no correctness as construction of VersionedValue is private.
if (previous.getZxid() != current.getZxid() || !Arrays.equals(previous.getValue(), current.getValue())) {
return false;
}
if (previous.getVersion() == -1) {
throw new IllegalTrySetVersionException();
}

try {
Stat result = client.setData().withVersion(previous.getVersion()).forPath(path, newValue);
updateValue(result.getVersion(), Arrays.copyOf(newValue, newValue.length));
updateValue(result.getMzxid(), result.getVersion(), Arrays.copyOf(newValue, newValue.length));
return true;
} catch (KeeperException.BadVersionException ignore) {
// ignore
Expand All @@ -193,14 +199,13 @@ public boolean trySetValue(VersionedValue<byte[]> previous, byte[] newValue) thr
return false;
}

private void updateValue(int version, byte[] bytes) {
private void updateValue(long zxid, int version, byte[] bytes) {
while (true) {
VersionedValue<byte[]> current = currentValue.get();
if (current.getVersion() >= version) {
// A newer version was concurrently set.
Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I am not sure wether SharedValue was designed to work with multiple owners, but I saw there is a background watcher to update value and also there is no rule to forbid concurrent usages. So, I assume it should work well in case of concurrency.

Then, let me assume a situation:

  1. current.getVersion is Integer.MAX_VALUE.
  2. Thread1 call trySetValue and succeed to get overflowed version Integer.MIN_VALUE, but the call to updateValue is somewhat delayed.
  3. Thread2 (assume watcher, which runs in ZooKeeper thread if I am not wrong) call updateVersion with version Integer.MIN_VALUE + 1. According to the code, this will be ignored.
  4. Thread1 call updateValue to continue its task with version Integer.MIN_VALUE. It succeeds.
  5. That is all, assume no changes anymore. I know it may not realistic.

current stores dated version while the javadoc says "All clients watching the same path will have the up-to-date value (considering ZK's normal consistency guarantees)".

Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I am not convinced in the overflow case either. @tisonkun @Hexiaoqiao

For the overflow case,

  • +1 to deprecate VersionedValue#getVersion so to warn clients about the "ordering assumption" if any about version.
  • +1 to a viable workaround if any and/or exception.

Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I think we should refactor updateValue a bit to compare ordering using Stat.mzxid. This way we are not fearing this overflow issue. I am stupid in reviwing without a deep thought, sorry for that. So my finally points are:

  • Deprecate VersionedValue#getVersion to warn clients about "ordering assumptions" and "overflow behavior".
  • Refactor updateValue to order using Stat.mzxid.
  • Throw exception in case of -1 Stat.version in trySetValue. I am positive to ZOOKEEPER-4743.
  • Document somehow about "overflow" and exception case in trySetValue.

Besides above, should we expose a VersionedValue#getZxid for client usage ?

Any thoughts @tisonkun @eolivelli @Hexiaoqiao ?

Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I have pushed new commits to go through above direction. Could you please take a look @Hexiaoqiao @tisonkun @eolivelli ?

if (current.getZxid() >= zxid) {
Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

What if only one client here and it goes through the overflow bound?

Said current.getZxid() == MAX_VALUE and zxid == MIN_VALUE, the update action will be skipped and the value will never updated.

Or we have different assumption on zxid?

Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

This is paranoid -:).

ZooKeeper guarantees a total order of messages, and it also guarantees a total order of proposals. ZooKeeper exposes the total ordering using a ZooKeeper transaction id (zxid). All proposals will be stamped with a zxid when it is proposed and exactly reflects the total ordering. -- https://zookeeper.apache.org/doc/r3.9.0/zookeeperInternals.html

Every change to the ZooKeeper state receives a stamp in the form of a zxid (ZooKeeper Transaction Id). This exposes the total ordering of all changes to ZooKeeper. Each change will have a unique zxid and if zxid1 is smaller than zxid2 then zxid1 happened before zxid2. -- https://zookeeper.apache.org/doc/r3.9.0/zookeeperProgrammers.html

In case of above situation, I believed that ZooKeeper is doomed to failure. The "never updated" should be negligible in case of the disaster.

return;
}
if (currentValue.compareAndSet(current, new VersionedValue<byte[]>(version, bytes))) {
if (currentValue.compareAndSet(current, new VersionedValue<>(zxid, version, bytes))) {
// Successfully set.
return;
}
Expand Down Expand Up @@ -248,14 +253,14 @@ private void readValue() throws Exception {
Stat localStat = new Stat();
byte[] bytes =
client.getData().storingStatIn(localStat).usingWatcher(watcher).forPath(path);
updateValue(localStat.getVersion(), bytes);
updateValue(localStat.getMzxid(), localStat.getVersion(), bytes);
}

private final BackgroundCallback upadateAndNotifyListenerCallback = new BackgroundCallback() {
@Override
public void processResult(CuratorFramework client, CuratorEvent event) throws Exception {
if (event.getResultCode() == KeeperException.Code.OK.intValue()) {
updateValue(event.getStat().getVersion(), event.getData());
updateValue(event.getStat().getMzxid(), event.getStat().getVersion(), event.getData());
notifyListeners();
}
}
Expand Down
Original file line number Diff line number Diff line change
Expand Up @@ -20,28 +20,47 @@
package org.apache.curator.framework.recipes.shared;

import com.google.common.base.Preconditions;
import java.util.function.Function;
import org.apache.zookeeper.data.Stat;

/**
* POJO for a version and a value
* POJO for versioned value.
*
* <p>Client must never construct this but get through {@link SharedValue#getVersionedValue()}
* or {@link SharedCount#getVersionedValue()}.
*/
public class VersionedValue<T> {
private final long zxid;
private final int version;
private final T value;

/**
* @param version the version
* @param value the value (cannot be null)
*/
VersionedValue(int version, T value) {
VersionedValue(long zxid, int version, T value) {
this.zxid = zxid;
this.version = version;
this.value = Preconditions.checkNotNull(value, "value cannot be null");
}

/**
* It is {@link Stat#getMzxid()} of the corresponding node.
*/
public long getZxid() {
return zxid;
}

/**
* It is {@link Stat#getVersion()} of the corresponding node.
*
* <p>It is known that this will overflow and hence not monotonic.
*/
public int getVersion() {
return version;
}

public T getValue() {
return value;
}

<R> VersionedValue<R> mapValue(Function<T, R> f) {
return new VersionedValue<>(zxid, version, f.apply(value));
}
}
Original file line number Diff line number Diff line change
Expand Up @@ -21,6 +21,7 @@

import static org.junit.jupiter.api.Assertions.assertEquals;
import static org.junit.jupiter.api.Assertions.assertFalse;
import static org.junit.jupiter.api.Assertions.assertThrows;
import static org.junit.jupiter.api.Assertions.assertTrue;
import com.google.common.collect.Lists;
import com.google.common.util.concurrent.ThreadFactoryBuilder;
Expand Down Expand Up @@ -51,6 +52,7 @@
import org.apache.zookeeper.WatchedEvent;
import org.junit.jupiter.api.Test;

@SuppressWarnings("deprecation")
public class TestSharedCount extends CuratorTestBase {
@Test
public void testMultiClients() throws Exception {
Expand Down Expand Up @@ -206,13 +208,20 @@ public void testSimpleVersioned() throws Exception {
assertEquals(count.getCount(), 10);

// Wrong value
assertFalse(count.trySetCount(new VersionedValue<Integer>(3, 20), 7));
assertFalse(count.trySetCount(new VersionedValue<>(current.getZxid(), 3, 20), 7));
// Wrong version
assertFalse(count.trySetCount(new VersionedValue<Integer>(10, 10), 7));
assertFalse(count.trySetCount(new VersionedValue<>(current.getZxid(), 10, 10), 7));
assertFalse(count.trySetCount(new VersionedValue<>(current.getZxid() + 1, 3, 10), 7));

// Server changed
client.setData().forPath("/count", SharedCount.toBytes(88));
assertFalse(count.trySetCount(current, 234));

assertThrows(IllegalTrySetVersionException.class, () -> {
VersionedValue<Integer> cached = count.getVersionedValue();
VersionedValue<Integer> illegal = new VersionedValue<>(cached.getZxid(), -1, cached.getValue());
count.trySetCount(illegal, 20);
});
} finally {
CloseableUtils.closeQuietly(count);
CloseableUtils.closeQuietly(client);
Expand Down