Line 0
Link Here
|
|
|
1 |
/* |
2 |
* Licensed to the Apache Software Foundation (ASF) under one or more |
3 |
* contributor license agreements. See the NOTICE file distributed with |
4 |
* this work for additional information regarding copyright ownership. |
5 |
* The ASF licenses this file to You under the Apache License, Version 2.0 |
6 |
* (the "License"); you may not use this file except in compliance with |
7 |
* the License. You may obtain a copy of the License at |
8 |
* |
9 |
* http://www.apache.org/licenses/LICENSE-2.0 |
10 |
* |
11 |
* Unless required by applicable law or agreed to in writing, software |
12 |
* distributed under the License is distributed on an "AS IS" BASIS, |
13 |
* WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. |
14 |
* See the License for the specific language governing permissions and |
15 |
* limitations under the License. |
16 |
*/ |
17 |
package org.apache.catalina.valves; |
18 |
|
19 |
import java.io.IOException; |
20 |
import java.util.Date; |
21 |
import java.util.Queue; |
22 |
import java.util.concurrent.ConcurrentHashMap; |
23 |
import java.util.concurrent.ConcurrentLinkedQueue; |
24 |
import java.util.concurrent.atomic.AtomicInteger; |
25 |
|
26 |
import javax.servlet.ServletException; |
27 |
import org.apache.catalina.Container; |
28 |
import org.apache.catalina.Engine; |
29 |
import org.apache.catalina.LifecycleException; |
30 |
|
31 |
import org.apache.catalina.connector.Request; |
32 |
import org.apache.catalina.connector.Response; |
33 |
import org.apache.juli.logging.Log; |
34 |
import org.apache.juli.logging.LogFactory; |
35 |
import org.apache.tomcat.util.res.StringManager; |
36 |
|
37 |
public class StuckThreadDetectionValve extends ValveBase { |
38 |
|
39 |
/** |
40 |
* The descriptive information related to this implementation. |
41 |
*/ |
42 |
private static final String info = |
43 |
"org.apache.catalina.valves.StuckThreadDetectionValve/1.0"; |
44 |
/** |
45 |
* Logger |
46 |
*/ |
47 |
private static final Log log = LogFactory.getLog(StuckThreadDetectionValve.class); |
48 |
/** |
49 |
* Keeps count of the number of stuck threads detected |
50 |
*/ |
51 |
private final AtomicInteger stuckCount = new AtomicInteger(0); |
52 |
/** |
53 |
* In seconds. Default 600 (10 minutes). |
54 |
*/ |
55 |
private int threshold = 600; |
56 |
/** |
57 |
* 'SO_TIMEOUT' is 20 seconds in Tomcat. |
58 |
* The threshold must at least be larger than that. |
59 |
*/ |
60 |
private static final int THRESHOLD_MIN_VALUE = 120; |
61 |
/** |
62 |
* The only references we keep to actual running Thread objects are in |
63 |
* this Map (which is automatically cleaned in invoke()s finally clause). |
64 |
* That way, Threads can be GC'ed, eventhough the Valve still thinks they |
65 |
* are stuck (caused by a long monitor interval) |
66 |
*/ |
67 |
private ConcurrentHashMap<Long, MonitoredThread> activeThreads = |
68 |
new ConcurrentHashMap<Long, MonitoredThread>(); |
69 |
/** |
70 |
* |
71 |
*/ |
72 |
private Queue<CompletedStuckThread> completedStuckThreadsQueue = |
73 |
new ConcurrentLinkedQueue<CompletedStuckThread>(); |
74 |
/** |
75 |
* The string manager for this package. |
76 |
*/ |
77 |
private static final StringManager sm = |
78 |
StringManager.getManager(Constants.Package); |
79 |
|
80 |
/** |
81 |
* Specify the threshold (in seconds) used when checking for stuck threads. |
82 |
* Minimum value is 120 seconds. The default is 600 seconds. |
83 |
* |
84 |
* @param threshold The new threshold in seconds |
85 |
*/ |
86 |
public void setThreshold(int threshold) { |
87 |
if (threshold >= THRESHOLD_MIN_VALUE) { |
88 |
this.threshold = threshold; |
89 |
if (log.isDebugEnabled()) { |
90 |
log.debug("Threshold set to " + threshold + " seconds"); |
91 |
} |
92 |
} else { |
93 |
log.warn("Threshold too low. Must be " |
94 |
+ THRESHOLD_MIN_VALUE |
95 |
+ " or larger. Actual: " + threshold); |
96 |
} |
97 |
} |
98 |
|
99 |
/** |
100 |
* @see #setThreshold(int) |
101 |
* @return The current threshold in seconds |
102 |
*/ |
103 |
public int getThreshold() { |
104 |
return threshold; |
105 |
} |
106 |
|
107 |
@Override |
108 |
protected void initInternal() throws LifecycleException { |
109 |
super.initInternal(); |
110 |
|
111 |
if (log.isDebugEnabled()) { |
112 |
log.debug("Monitoring stuck threads with threshold = " |
113 |
+ threshold |
114 |
+ " sec"); |
115 |
} |
116 |
} |
117 |
|
118 |
/** |
119 |
* Return descriptive information about this Valve implementation. |
120 |
*/ |
121 |
@Override |
122 |
public String getInfo() { |
123 |
return info; |
124 |
} |
125 |
|
126 |
private void notifyStuckThreadDetected(String threadName, |
127 |
StackTraceElement[] trace, long activeTime, |
128 |
Date startTime, int numStuckThreads) { |
129 |
|
130 |
String msg = sm.getString( |
131 |
"stuckThreadDetectionValve.notifyStuckThreadDetected", |
132 |
threadName, activeTime, startTime, numStuckThreads); |
133 |
msg += "\n" + getStackTraceAsString(trace); |
134 |
log.warn(msg); |
135 |
} |
136 |
|
137 |
private void notifyStuckThreadCompleted(String threadName, |
138 |
long activeTime, int numStuckThreads) { |
139 |
|
140 |
String msg = sm.getString( |
141 |
"stuckThreadDetectionValve.notifyStuckThreadCompleted", |
142 |
threadName, activeTime, numStuckThreads); |
143 |
//Since the "stuck thread notification" is warn, this should also be warn |
144 |
log.warn(msg); |
145 |
} |
146 |
|
147 |
//TODO Utility method, should be moved to utility class |
148 |
public static String getStackTraceAsString(StackTraceElement[] trace) { |
149 |
StringBuilder buf = new StringBuilder(); |
150 |
for (int i = 0; i < trace.length; i++) { |
151 |
buf.append("\tat ").append(trace[i]).append("\n"); |
152 |
} |
153 |
return buf.toString(); |
154 |
} |
155 |
|
156 |
//FIXME where does it make sense to add this Valve? |
157 |
/** |
158 |
* Set the Container to which this Valve is attached. |
159 |
* |
160 |
* @param container The container to which we are attached |
161 |
*/ |
162 |
@Override |
163 |
public void setContainer(Container container) { |
164 |
|
165 |
if (container == null){ |
166 |
throw new IllegalArgumentException("Configuration error: " |
167 |
+ "Container cannot be null. Valve must be attached to an Engine"); |
168 |
} |
169 |
if(!(container instanceof Engine)){ |
170 |
throw new IllegalArgumentException("Configuration error: " |
171 |
+ "Valve must be attached to an Engine. Actual: " |
172 |
+container.getClass().getName()); |
173 |
} |
174 |
|
175 |
super.setContainer(container); |
176 |
// this.engine = (Engine) container; |
177 |
} |
178 |
|
179 |
/** |
180 |
* {@inheritDoc} |
181 |
*/ |
182 |
@Override |
183 |
public void invoke(Request request, Response response) |
184 |
throws IOException, ServletException { |
185 |
//Save the thread/runnable |
186 |
//Keeping a reference to the thread object here does not prevent GC'ing, |
187 |
//as the reference is removed from the Map in the finally clause |
188 |
|
189 |
Long key = new Long(Thread.currentThread().getId()); |
190 |
|
191 |
MonitoredThread previous = activeThreads.put(key, |
192 |
new MonitoredThread(Thread.currentThread())); |
193 |
|
194 |
//In theory we might get collisions when using "new Long(thread.getId()).hashCode()". |
195 |
//We can prevent this by using new Integer(AtomicInt.getAndIncrement()) for uniqueness |
196 |
if (previous != null) { |
197 |
log.error("Key already added to map. Previous value has been removed (" |
198 |
+ previous.getThread().getName() + ")"); |
199 |
} |
200 |
|
201 |
if (log.isTraceEnabled()) { |
202 |
log.trace("Monitoring execution time for Request " |
203 |
+ request.getRequestURI() + " with originalRemoteAddr '" |
204 |
+ request.getRemoteAddr() + ":" + request.getRemotePort() |
205 |
+ "'"); |
206 |
} |
207 |
|
208 |
try { |
209 |
getNext().invoke(request, response); |
210 |
} finally { |
211 |
MonitoredThread monitoredThread = activeThreads.remove(key); |
212 |
if (monitoredThread.isMarkedAsStuck()) { |
213 |
completedStuckThreadsQueue.add( |
214 |
new CompletedStuckThread(monitoredThread.getThread().getName(), |
215 |
monitoredThread.getActiveTimeInMillis())); |
216 |
} |
217 |
} |
218 |
} |
219 |
|
220 |
@Override |
221 |
public void backgroundProcess() { |
222 |
super.backgroundProcess(); |
223 |
|
224 |
long thresholdInMillis = threshold * 1000; |
225 |
|
226 |
//Check monitored threads |
227 |
for (MonitoredThread monitoredThread : activeThreads.values()) { |
228 |
long activeTime = monitoredThread.getActiveTimeInMillis(); |
229 |
|
230 |
if (!monitoredThread.isMarkedAsStuck() && activeTime >= thresholdInMillis) { |
231 |
int numStuckThreads = stuckCount.incrementAndGet(); |
232 |
monitoredThread.markAsStuck(); |
233 |
notifyStuckThreadDetected(monitoredThread.getThread().getName(), |
234 |
monitoredThread.getThread().getStackTrace(), |
235 |
activeTime, monitoredThread.getStartTime(), numStuckThreads); |
236 |
} |
237 |
} |
238 |
//Check if any threads previously reported as stuck, have finished. |
239 |
CompletedStuckThread completedStuckThread = completedStuckThreadsQueue.poll(); |
240 |
while (completedStuckThread != null) { |
241 |
int numStuckThreads = stuckCount.decrementAndGet(); |
242 |
notifyStuckThreadCompleted(completedStuckThread.getName(), |
243 |
completedStuckThread.getTotalActiveTime(), numStuckThreads); |
244 |
completedStuckThread = completedStuckThreadsQueue.poll(); |
245 |
} |
246 |
} |
247 |
|
248 |
private class MonitoredThread { |
249 |
|
250 |
/** |
251 |
* Reference to the thread to get a stack trace from background task |
252 |
*/ |
253 |
private Thread thread; |
254 |
private long start = System.currentTimeMillis(); |
255 |
private volatile boolean isStuck = false; |
256 |
|
257 |
public MonitoredThread(Thread thread) { |
258 |
this.thread = thread; |
259 |
} |
260 |
|
261 |
public Thread getThread() { |
262 |
return this.thread; |
263 |
} |
264 |
|
265 |
public long getActiveTimeInMillis() { |
266 |
return System.currentTimeMillis() - start; |
267 |
} |
268 |
|
269 |
public Date getStartTime() { |
270 |
return new Date(start); |
271 |
} |
272 |
|
273 |
public void markAsStuck() { |
274 |
this.isStuck = true; |
275 |
} |
276 |
|
277 |
public boolean isMarkedAsStuck() { |
278 |
return this.isStuck; |
279 |
} |
280 |
} |
281 |
|
282 |
private class CompletedStuckThread { |
283 |
|
284 |
private String threadName; |
285 |
private long totalActiveTime; |
286 |
|
287 |
public CompletedStuckThread(String threadName, long totalActiveTime) { |
288 |
this.threadName = threadName; |
289 |
this.totalActiveTime = totalActiveTime; |
290 |
} |
291 |
|
292 |
public String getName() { |
293 |
return this.threadName; |
294 |
} |
295 |
|
296 |
public long getTotalActiveTime() { |
297 |
return this.totalActiveTime; |
298 |
} |
299 |
} |
300 |
} |