001/* 002 * Licensed to the Apache Software Foundation (ASF) under one 003 * or more contributor license agreements. See the NOTICE file 004 * distributed with this work for additional information 005 * regarding copyright ownership. The ASF licenses this file 006 * to you under the Apache License, Version 2.0 (the 007 * "License"); you may not use this file except in compliance 008 * with the License. You may obtain a copy of the License at 009 * 010 * http://www.apache.org/licenses/LICENSE-2.0 011 * 012 * Unless required by applicable law or agreed to in writing, software 013 * distributed under the License is distributed on an "AS IS" BASIS, 014 * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. 015 * See the License for the specific language governing permissions and 016 * limitations under the License. 017 */ 018package org.apache.hadoop.hbase.http; 019 020import java.io.File; 021import java.io.IOException; 022import java.util.ArrayList; 023import java.util.List; 024import java.util.concurrent.TimeUnit; 025import java.util.concurrent.atomic.AtomicInteger; 026import java.util.concurrent.locks.Lock; 027import java.util.concurrent.locks.ReentrantLock; 028import javax.servlet.http.HttpServlet; 029import javax.servlet.http.HttpServletRequest; 030import javax.servlet.http.HttpServletResponse; 031import org.apache.hadoop.hbase.util.ProcessUtils; 032import org.apache.yetus.audience.InterfaceAudience; 033import org.slf4j.Logger; 034import org.slf4j.LoggerFactory; 035 036import org.apache.hbase.thirdparty.com.google.common.base.Joiner; 037 038/** 039 * Servlet that runs async-profiler as web-endpoint. Following options from async-profiler can be 040 * specified as query paramater. // -e event profiling event: cpu|alloc|lock|cache-misses etc. // -d 041 * duration run profiling for 'duration' seconds (integer) // -i interval sampling interval in 042 * nanoseconds (long) // -j jstackdepth maximum Java stack depth (integer) // -b bufsize frame 043 * buffer size (long) // -t profile different threads separately // -s simple class names instead of 044 * FQN // -o fmt[,fmt...] output format: summary|traces|flat|collapsed|svg|tree|jfr|html // --width 045 * px SVG width pixels (integer) // --height px SVG frame height pixels (integer) // --minwidth px 046 * skip frames smaller than px (double) // --reverse generate stack-reversed FlameGraph / Call tree 047 * Example: - To collect 30 second CPU profile of current process (returns FlameGraph svg) curl 048 * "http://localhost:10002/prof" - To collect 1 minute CPU profile of current process and output in 049 * tree format (html) curl "http://localhost:10002/prof?output=tree&duration=60" - To collect 30 050 * second heap allocation profile of current process (returns FlameGraph svg) curl 051 * "http://localhost:10002/prof?event=alloc" - To collect lock contention profile of current process 052 * (returns FlameGraph svg) curl "http://localhost:10002/prof?event=lock" Following event types are 053 * supported (default is 'cpu') (NOTE: not all OS'es support all events) // Perf events: // cpu // 054 * page-faults // context-switches // cycles // instructions // cache-references // cache-misses // 055 * branches // branch-misses // bus-cycles // L1-dcache-load-misses // LLC-load-misses // 056 * dTLB-load-misses // mem:breakpoint // trace:tracepoint // Java events: // alloc // lock 057 */ 058@InterfaceAudience.Private 059public class ProfileServlet extends HttpServlet { 060 061 private static final long serialVersionUID = 1L; 062 private static final Logger LOG = LoggerFactory.getLogger(ProfileServlet.class); 063 064 private static final String ACCESS_CONTROL_ALLOW_METHODS = "Access-Control-Allow-Methods"; 065 private static final String ALLOWED_METHODS = "GET"; 066 private static final String ACCESS_CONTROL_ALLOW_ORIGIN = "Access-Control-Allow-Origin"; 067 private static final String CONTENT_TYPE_TEXT = "text/plain; charset=utf-8"; 068 private static final String ASYNC_PROFILER_HOME_ENV = "ASYNC_PROFILER_HOME"; 069 private static final String ASYNC_PROFILER_HOME_SYSTEM_PROPERTY = "async.profiler.home"; 070 private static final String PROFILER_SCRIPT = "/profiler.sh"; 071 private static final int DEFAULT_DURATION_SECONDS = 10; 072 private static final AtomicInteger ID_GEN = new AtomicInteger(0); 073 static final String OUTPUT_DIR = System.getProperty("java.io.tmpdir") + "/prof-output-hbase"; 074 075 enum Event { 076 CPU("cpu"), 077 WALL("wall"), 078 ALLOC("alloc"), 079 LOCK("lock"), 080 PAGE_FAULTS("page-faults"), 081 CONTEXT_SWITCHES("context-switches"), 082 CYCLES("cycles"), 083 INSTRUCTIONS("instructions"), 084 CACHE_REFERENCES("cache-references"), 085 CACHE_MISSES("cache-misses"), 086 BRANCHES("branches"), 087 BRANCH_MISSES("branch-misses"), 088 BUS_CYCLES("bus-cycles"), 089 L1_DCACHE_LOAD_MISSES("L1-dcache-load-misses"), 090 LLC_LOAD_MISSES("LLC-load-misses"), 091 DTLB_LOAD_MISSES("dTLB-load-misses"), 092 MEM_BREAKPOINT("mem:breakpoint"), 093 TRACE_TRACEPOINT("trace:tracepoint"),; 094 095 private final String internalName; 096 097 Event(final String internalName) { 098 this.internalName = internalName; 099 } 100 101 public String getInternalName() { 102 return internalName; 103 } 104 105 public static Event fromInternalName(final String name) { 106 for (Event event : values()) { 107 if (event.getInternalName().equalsIgnoreCase(name)) { 108 return event; 109 } 110 } 111 112 return null; 113 } 114 } 115 116 enum Output { 117 SUMMARY, 118 TRACES, 119 FLAT, 120 COLLAPSED, 121 // No SVG in 2.x asyncprofiler. 122 SVG, 123 TREE, 124 JFR, 125 // In 2.x asyncprofiler, this is how you get flamegraphs. 126 HTML 127 } 128 129 @edu.umd.cs.findbugs.annotations.SuppressWarnings(value = "SE_TRANSIENT_FIELD_NOT_RESTORED", 130 justification = "This class is never serialized nor restored.") 131 private transient Lock profilerLock = new ReentrantLock(); 132 private transient volatile Process process; 133 private String asyncProfilerHome; 134 private Integer pid; 135 136 public ProfileServlet() { 137 this.asyncProfilerHome = getAsyncProfilerHome(); 138 this.pid = ProcessUtils.getPid(); 139 LOG.info("Servlet process PID: " + pid + " asyncProfilerHome: " + asyncProfilerHome); 140 } 141 142 @Override 143 protected void doGet(final HttpServletRequest req, final HttpServletResponse resp) 144 throws IOException { 145 if (!HttpServer.isInstrumentationAccessAllowed(getServletContext(), req, resp)) { 146 resp.setStatus(HttpServletResponse.SC_UNAUTHORIZED); 147 setResponseHeader(resp); 148 resp.getWriter().write("Unauthorized: Instrumentation access is not allowed!"); 149 return; 150 } 151 152 // make sure async profiler home is set 153 if (asyncProfilerHome == null || asyncProfilerHome.trim().isEmpty()) { 154 resp.setStatus(HttpServletResponse.SC_INTERNAL_SERVER_ERROR); 155 setResponseHeader(resp); 156 resp.getWriter() 157 .write("ASYNC_PROFILER_HOME env is not set.\n\n" 158 + "Please ensure the prerequsites for the Profiler Servlet have been installed and the\n" 159 + "environment is properly configured. For more information please see\n" 160 + "http://hbase.apache.org/book.html#profiler\n"); 161 return; 162 } 163 164 // if pid is explicitly specified, use it else default to current process 165 pid = getInteger(req, "pid", pid); 166 167 // if pid is not specified in query param and if current process pid cannot be determined 168 if (pid == null) { 169 resp.setStatus(HttpServletResponse.SC_INTERNAL_SERVER_ERROR); 170 setResponseHeader(resp); 171 resp.getWriter() 172 .write("'pid' query parameter unspecified or unable to determine PID of current process."); 173 return; 174 } 175 176 final int duration = getInteger(req, "duration", DEFAULT_DURATION_SECONDS); 177 final Output output = getOutput(req); 178 final Event event = getEvent(req); 179 final Long interval = getLong(req, "interval"); 180 final Integer jstackDepth = getInteger(req, "jstackdepth", null); 181 final Long bufsize = getLong(req, "bufsize"); 182 final boolean thread = req.getParameterMap().containsKey("thread"); 183 final boolean simple = req.getParameterMap().containsKey("simple"); 184 final Integer width = getInteger(req, "width", null); 185 final Integer height = getInteger(req, "height", null); 186 final Double minwidth = getMinWidth(req); 187 final boolean reverse = req.getParameterMap().containsKey("reverse"); 188 189 if (process == null || !process.isAlive()) { 190 try { 191 int lockTimeoutSecs = 3; 192 if (profilerLock.tryLock(lockTimeoutSecs, TimeUnit.SECONDS)) { 193 try { 194 File outputFile = 195 new File(OUTPUT_DIR, "async-prof-pid-" + pid + "-" + event.name().toLowerCase() + "-" 196 + ID_GEN.incrementAndGet() + "." + output.name().toLowerCase()); 197 List<String> cmd = new ArrayList<>(); 198 cmd.add(asyncProfilerHome + PROFILER_SCRIPT); 199 cmd.add("-e"); 200 cmd.add(event.getInternalName()); 201 cmd.add("-d"); 202 cmd.add("" + duration); 203 cmd.add("-o"); 204 cmd.add(output.name().toLowerCase()); 205 cmd.add("-f"); 206 cmd.add(outputFile.getAbsolutePath()); 207 if (interval != null) { 208 cmd.add("-i"); 209 cmd.add(interval.toString()); 210 } 211 if (jstackDepth != null) { 212 cmd.add("-j"); 213 cmd.add(jstackDepth.toString()); 214 } 215 if (bufsize != null) { 216 cmd.add("-b"); 217 cmd.add(bufsize.toString()); 218 } 219 if (thread) { 220 cmd.add("-t"); 221 } 222 if (simple) { 223 cmd.add("-s"); 224 } 225 if (width != null) { 226 cmd.add("--width"); 227 cmd.add(width.toString()); 228 } 229 if (height != null) { 230 cmd.add("--height"); 231 cmd.add(height.toString()); 232 } 233 if (minwidth != null) { 234 cmd.add("--minwidth"); 235 cmd.add(minwidth.toString()); 236 } 237 if (reverse) { 238 cmd.add("--reverse"); 239 } 240 cmd.add(pid.toString()); 241 process = ProcessUtils.runCmdAsync(cmd); 242 243 // set response and set refresh header to output location 244 setResponseHeader(resp); 245 resp.setStatus(HttpServletResponse.SC_ACCEPTED); 246 String relativeUrl = "/prof-output-hbase/" + outputFile.getName(); 247 resp.getWriter() 248 .write("Started [" + event.getInternalName() 249 + "] profiling. This page will automatically redirect to " + relativeUrl + " after " 250 + duration + " seconds. " 251 + "If empty diagram and Linux 4.6+, see 'Basic Usage' section on the Async " 252 + "Profiler Home Page, https://github.com/jvm-profiling-tools/async-profiler." 253 + "\n\nCommand:\n" + Joiner.on(" ").join(cmd)); 254 255 // to avoid auto-refresh by ProfileOutputServlet, refreshDelay can be specified 256 // via url param 257 int refreshDelay = getInteger(req, "refreshDelay", 0); 258 259 // instead of sending redirect, set auto-refresh so that browsers will refresh 260 // with redirected url 261 resp.setHeader("Refresh", (duration + refreshDelay) + ";" + relativeUrl); 262 resp.getWriter().flush(); 263 } finally { 264 profilerLock.unlock(); 265 } 266 } else { 267 setResponseHeader(resp); 268 resp.setStatus(HttpServletResponse.SC_INTERNAL_SERVER_ERROR); 269 resp.getWriter() 270 .write("Unable to acquire lock. Another instance of profiler might be running."); 271 LOG.warn("Unable to acquire lock in " + lockTimeoutSecs 272 + " seconds. Another instance of profiler might be running."); 273 } 274 } catch (InterruptedException e) { 275 LOG.warn("Interrupted while acquiring profile lock.", e); 276 resp.setStatus(HttpServletResponse.SC_INTERNAL_SERVER_ERROR); 277 } 278 } else { 279 setResponseHeader(resp); 280 resp.setStatus(HttpServletResponse.SC_INTERNAL_SERVER_ERROR); 281 resp.getWriter().write("Another instance of profiler is already running."); 282 } 283 } 284 285 private Integer getInteger(final HttpServletRequest req, final String param, 286 final Integer defaultValue) { 287 final String value = req.getParameter(param); 288 if (value != null) { 289 try { 290 return Integer.valueOf(value); 291 } catch (NumberFormatException e) { 292 return defaultValue; 293 } 294 } 295 return defaultValue; 296 } 297 298 private Long getLong(final HttpServletRequest req, final String param) { 299 final String value = req.getParameter(param); 300 if (value != null) { 301 try { 302 return Long.valueOf(value); 303 } catch (NumberFormatException e) { 304 return null; 305 } 306 } 307 return null; 308 } 309 310 private Double getMinWidth(final HttpServletRequest req) { 311 final String value = req.getParameter("minwidth"); 312 if (value != null) { 313 try { 314 return Double.valueOf(value); 315 } catch (NumberFormatException e) { 316 return null; 317 } 318 } 319 return null; 320 } 321 322 private Event getEvent(final HttpServletRequest req) { 323 final String eventArg = req.getParameter("event"); 324 if (eventArg != null) { 325 Event event = Event.fromInternalName(eventArg); 326 return event == null ? Event.CPU : event; 327 } 328 return Event.CPU; 329 } 330 331 private Output getOutput(final HttpServletRequest req) { 332 final String outputArg = req.getParameter("output"); 333 if (req.getParameter("output") != null) { 334 try { 335 return Output.valueOf(outputArg.trim().toUpperCase()); 336 } catch (IllegalArgumentException e) { 337 return Output.HTML; 338 } 339 } 340 return Output.HTML; 341 } 342 343 static void setResponseHeader(final HttpServletResponse response) { 344 response.setHeader(ACCESS_CONTROL_ALLOW_METHODS, ALLOWED_METHODS); 345 response.setHeader(ACCESS_CONTROL_ALLOW_ORIGIN, "*"); 346 response.setContentType(CONTENT_TYPE_TEXT); 347 } 348 349 static String getAsyncProfilerHome() { 350 String asyncProfilerHome = System.getenv(ASYNC_PROFILER_HOME_ENV); 351 // if ENV is not set, see if -Dasync.profiler.home=/path/to/async/profiler/home is set 352 if (asyncProfilerHome == null || asyncProfilerHome.trim().isEmpty()) { 353 asyncProfilerHome = System.getProperty(ASYNC_PROFILER_HOME_SYSTEM_PROPERTY); 354 } 355 356 return asyncProfilerHome; 357 } 358 359 public static class DisabledServlet extends HttpServlet { 360 361 private static final long serialVersionUID = 1L; 362 363 @Override 364 protected void doGet(final HttpServletRequest req, final HttpServletResponse resp) 365 throws IOException { 366 resp.setStatus(HttpServletResponse.SC_INTERNAL_SERVER_ERROR); 367 setResponseHeader(resp); 368 resp.getWriter() 369 .write("The profiler servlet was disabled at startup.\n\n" 370 + "Please ensure the prerequisites for the Profiler Servlet have been installed and the\n" 371 + "environment is properly configured. For more information please see\n" 372 + "http://hbase.apache.org/book.html#profiler\n"); 373 return; 374 } 375 376 } 377 378}