1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
|
// Copyright 2017 Yahoo Holdings. Licensed under the terms of the Apache 2.0 license. See LICENSE in the project root.
package com.yahoo.search.cluster;
import com.yahoo.concurrent.ThreadFactoryFactory;
import com.yahoo.search.result.ErrorMessage;
import java.util.Collections;
import java.util.Iterator;
import java.util.List;
import java.util.Map;
import java.util.concurrent.Executor;
import java.util.concurrent.ExecutorService;
import java.util.concurrent.Executors;
import java.util.concurrent.TimeUnit;
import java.util.concurrent.atomic.AtomicBoolean;
import java.util.logging.Level;
import java.util.logging.Logger;
/**
* Monitors of a cluster of remote nodes.
* The monitor uses an internal thread for node monitoring.
* All <i>public</i> methods of this class are multithread safe.
*
* @author bratseth
*/
public class ClusterMonitor<T> {
private final MonitorConfiguration configuration = new MonitorConfiguration();
private static Logger log = Logger.getLogger(ClusterMonitor.class.getName());
private final NodeManager<T> nodeManager;
private final MonitorThread monitorThread;
private final AtomicBoolean closed = new AtomicBoolean(false);
/** A map from Node to corresponding MonitoredNode */
private final Map<T, TrafficNodeMonitor<T>> nodeMonitors = Collections.synchronizedMap(new java.util.LinkedHashMap<>());
/** @deprecated It is not advised to start the monitoring thread in the constructor.
* Use ClusterMonitor(NodeManager manager, false) and explicit start(). */
@Deprecated
public ClusterMonitor(NodeManager<T> manager) {
this(manager, true);
}
public ClusterMonitor(NodeManager<T> manager, boolean startPingThread) {
nodeManager = manager;
monitorThread = new MonitorThread("search.clustermonitor");
if (startPingThread) {
monitorThread.start();
}
}
public void start() {
if ( ! monitorThread.isAlive()) {
monitorThread.start();
}
}
/** Returns the configuration of this cluster monitor */
public MonitorConfiguration getConfiguration() { return configuration; }
/**
* Adds a new node for monitoring.
* The object representing the node must
* <ul>
* <li>Have a sensible toString</li>
* <li>Have a sensible identity (equals and hashCode)</li>
* </ul>
*
* @param node the object representing the node
* @param internal whether or not this node is internal to this cluster
*/
public void add(T node, boolean internal) {
nodeMonitors.put(node, new TrafficNodeMonitor<>(node, configuration, internal));
}
/**
* Returns the monitor of the given node, or null if this node has not been added
*/
public BaseNodeMonitor<T> getNodeMonitor(T node) {
return nodeMonitors.get(node);
}
/** Called from ClusterSearcher/NodeManager when a node failed */
public synchronized void failed(T node, ErrorMessage error) {
if (closed.get()) return; // Do not touch state if close has started.
TrafficNodeMonitor<T> monitor = nodeMonitors.get(node);
Boolean wasWorking = monitor.isKnownWorking();
monitor.failed(error);
if (wasWorking != monitor.isKnownWorking())
nodeManager.failed(node);
}
/** Called when a node responded */
public synchronized void responded(T node) {
if (closed.get()) return; // Do not touch state if close has started.
TrafficNodeMonitor<T> monitor = nodeMonitors.get(node);
Boolean wasWorking = monitor.isKnownWorking();
monitor.responded();
if (wasWorking != monitor.isKnownWorking())
nodeManager.working(monitor.getNode());
}
/**
* Ping all nodes which needs pinging to discover state changes
*/
public void ping(Executor executor) {
for (Iterator<BaseNodeMonitor<T>> i = nodeMonitorIterator(); i.hasNext() && !closed.get(); ) {
BaseNodeMonitor<T> monitor= i.next();
nodeManager.ping(this, monitor.getNode(), executor); // Cause call to failed or responded
}
if (closed.get()) return; // Do nothing to change state if close has started.
nodeManager.pingIterationCompleted();
}
/** Returns a thread-safe snapshot of the NodeMonitors of all added nodes */
public Iterator<BaseNodeMonitor<T>> nodeMonitorIterator() {
return nodeMonitors().iterator();
}
/** Returns a thread-safe snapshot of the NodeMonitors of all added nodes */
public List<BaseNodeMonitor<T>> nodeMonitors() {
return new java.util.ArrayList<>(nodeMonitors.values());
}
/** Must be called when this goes out of use */
public void shutdown() {
closed.set(true);
synchronized (this) {
nodeMonitors.clear();
}
synchronized (nodeManager) {
nodeManager.notifyAll();
}
try {
if (monitorThread.isAlive()) {
monitorThread.join();
}
} catch (InterruptedException e) {}
}
private class MonitorThread extends Thread {
MonitorThread(String name) {
super(name);
}
public void run() {
log.info("Starting cluster monitor thread " + getName());
// Pings must happen in a separate thread from this to handle timeouts
// By using a cached thread pool we ensured that 1) a single thread will be used
// for all pings when there are no problems (important because it ensures that
// any thread local connections are reused) 2) a new thread will be started to execute
// new pings when a ping is not responding
ExecutorService pingExecutor=Executors.newCachedThreadPool(ThreadFactoryFactory.getDaemonThreadFactory("search.ping"));
while (!closed.get()) {
try {
log.finest("Activating ping");
ping(pingExecutor);
synchronized (nodeManager) {
nodeManager.wait(configuration.getCheckInterval());
}
}
catch (Throwable e) {
if (closed.get() && e instanceof InterruptedException) {
break;
} else if ( ! (e instanceof Exception) ) {
log.log(Level.WARNING,"Error in monitor thread, will quit", e);
break;
} else {
log.log(Level.WARNING,"Exception in monitor thread", e);
}
}
}
pingExecutor.shutdown();
try {
pingExecutor.awaitTermination(10, TimeUnit.SECONDS);
} catch (InterruptedException e) { }
log.info("Stopped cluster monitor thread " + getName());
}
}
}
|
