1 | <!DOCTYPE html PUBLIC "-//W3C//DTD HTML 4.01 Transitional//EN" "http://www.w3.org/TR/html4/loose.dtd"> |
---|
2 | <html> |
---|
3 | <head> |
---|
4 | <META http-equiv="Content-Type" content="text/html; charset=UTF-8"> |
---|
5 | <meta content="Apache Forrest" name="Generator"> |
---|
6 | <meta name="Forrest-version" content="0.8"> |
---|
7 | <meta name="Forrest-skin-name" content="pelt"> |
---|
8 | <title>Fair Scheduler Guide</title> |
---|
9 | <link type="text/css" href="skin/basic.css" rel="stylesheet"> |
---|
10 | <link media="screen" type="text/css" href="skin/screen.css" rel="stylesheet"> |
---|
11 | <link media="print" type="text/css" href="skin/print.css" rel="stylesheet"> |
---|
12 | <link type="text/css" href="skin/profile.css" rel="stylesheet"> |
---|
13 | <script src="skin/getBlank.js" language="javascript" type="text/javascript"></script><script src="skin/getMenu.js" language="javascript" type="text/javascript"></script><script src="skin/fontsize.js" language="javascript" type="text/javascript"></script> |
---|
14 | <link rel="shortcut icon" href="images/favicon.ico"> |
---|
15 | </head> |
---|
16 | <body onload="init()"> |
---|
17 | <script type="text/javascript">ndeSetTextSize();</script> |
---|
18 | <div id="top"> |
---|
19 | <!--+ |
---|
20 | |breadtrail |
---|
21 | +--> |
---|
22 | <div class="breadtrail"> |
---|
23 | <a href="http://www.apache.org/">Apache</a> > <a href="http://hadoop.apache.org/">Hadoop</a> > <a href="http://hadoop.apache.org/core/">Core</a><script src="skin/breadcrumbs.js" language="JavaScript" type="text/javascript"></script> |
---|
24 | </div> |
---|
25 | <!--+ |
---|
26 | |header |
---|
27 | +--> |
---|
28 | <div class="header"> |
---|
29 | <!--+ |
---|
30 | |start group logo |
---|
31 | +--> |
---|
32 | <div class="grouplogo"> |
---|
33 | <a href="http://hadoop.apache.org/"><img class="logoImage" alt="Hadoop" src="images/hadoop-logo.jpg" title="Apache Hadoop"></a> |
---|
34 | </div> |
---|
35 | <!--+ |
---|
36 | |end group logo |
---|
37 | +--> |
---|
38 | <!--+ |
---|
39 | |start Project Logo |
---|
40 | +--> |
---|
41 | <div class="projectlogo"> |
---|
42 | <a href="http://hadoop.apache.org/core/"><img class="logoImage" alt="Hadoop" src="images/core-logo.gif" title="Scalable Computing Platform"></a> |
---|
43 | </div> |
---|
44 | <!--+ |
---|
45 | |end Project Logo |
---|
46 | +--> |
---|
47 | <!--+ |
---|
48 | |start Search |
---|
49 | +--> |
---|
50 | <div class="searchbox"> |
---|
51 | <form action="http://www.google.com/search" method="get" class="roundtopsmall"> |
---|
52 | <input value="hadoop.apache.org" name="sitesearch" type="hidden"><input onFocus="getBlank (this, 'Search the site with google');" size="25" name="q" id="query" type="text" value="Search the site with google"> |
---|
53 | <input name="Search" value="Search" type="submit"> |
---|
54 | </form> |
---|
55 | </div> |
---|
56 | <!--+ |
---|
57 | |end search |
---|
58 | +--> |
---|
59 | <!--+ |
---|
60 | |start Tabs |
---|
61 | +--> |
---|
62 | <ul id="tabs"> |
---|
63 | <li> |
---|
64 | <a class="unselected" href="http://hadoop.apache.org/core/">Project</a> |
---|
65 | </li> |
---|
66 | <li> |
---|
67 | <a class="unselected" href="http://wiki.apache.org/hadoop">Wiki</a> |
---|
68 | </li> |
---|
69 | <li class="current"> |
---|
70 | <a class="selected" href="index.html">Hadoop 0.20 Documentation</a> |
---|
71 | </li> |
---|
72 | </ul> |
---|
73 | <!--+ |
---|
74 | |end Tabs |
---|
75 | +--> |
---|
76 | </div> |
---|
77 | </div> |
---|
78 | <div id="main"> |
---|
79 | <div id="publishedStrip"> |
---|
80 | <!--+ |
---|
81 | |start Subtabs |
---|
82 | +--> |
---|
83 | <div id="level2tabs"></div> |
---|
84 | <!--+ |
---|
85 | |end Endtabs |
---|
86 | +--> |
---|
87 | <script type="text/javascript"><!-- |
---|
88 | document.write("Last Published: " + document.lastModified); |
---|
89 | // --></script> |
---|
90 | </div> |
---|
91 | <!--+ |
---|
92 | |breadtrail |
---|
93 | +--> |
---|
94 | <div class="breadtrail"> |
---|
95 | |
---|
96 | |
---|
97 | </div> |
---|
98 | <!--+ |
---|
99 | |start Menu, mainarea |
---|
100 | +--> |
---|
101 | <!--+ |
---|
102 | |start Menu |
---|
103 | +--> |
---|
104 | <div id="menu"> |
---|
105 | <div onclick="SwitchMenu('menu_1.1', 'skin/')" id="menu_1.1Title" class="menutitle">Getting Started</div> |
---|
106 | <div id="menu_1.1" class="menuitemgroup"> |
---|
107 | <div class="menuitem"> |
---|
108 | <a href="index.html">Overview</a> |
---|
109 | </div> |
---|
110 | <div class="menuitem"> |
---|
111 | <a href="quickstart.html">Quick Start</a> |
---|
112 | </div> |
---|
113 | <div class="menuitem"> |
---|
114 | <a href="cluster_setup.html">Cluster Setup</a> |
---|
115 | </div> |
---|
116 | <div class="menuitem"> |
---|
117 | <a href="mapred_tutorial.html">Map/Reduce Tutorial</a> |
---|
118 | </div> |
---|
119 | </div> |
---|
120 | <div onclick="SwitchMenu('menu_selected_1.2', 'skin/')" id="menu_selected_1.2Title" class="menutitle" style="background-image: url('skin/images/chapter_open.gif');">Programming Guides</div> |
---|
121 | <div id="menu_selected_1.2" class="selectedmenuitemgroup" style="display: block;"> |
---|
122 | <div class="menuitem"> |
---|
123 | <a href="commands_manual.html">Commands</a> |
---|
124 | </div> |
---|
125 | <div class="menuitem"> |
---|
126 | <a href="distcp.html">DistCp</a> |
---|
127 | </div> |
---|
128 | <div class="menuitem"> |
---|
129 | <a href="native_libraries.html">Native Libraries</a> |
---|
130 | </div> |
---|
131 | <div class="menuitem"> |
---|
132 | <a href="streaming.html">Streaming</a> |
---|
133 | </div> |
---|
134 | <div class="menupage"> |
---|
135 | <div class="menupagetitle">Fair Scheduler</div> |
---|
136 | </div> |
---|
137 | <div class="menuitem"> |
---|
138 | <a href="capacity_scheduler.html">Capacity Scheduler</a> |
---|
139 | </div> |
---|
140 | <div class="menuitem"> |
---|
141 | <a href="service_level_auth.html">Service Level Authorization</a> |
---|
142 | </div> |
---|
143 | <div class="menuitem"> |
---|
144 | <a href="vaidya.html">Vaidya</a> |
---|
145 | </div> |
---|
146 | <div class="menuitem"> |
---|
147 | <a href="hadoop_archives.html">Archives</a> |
---|
148 | </div> |
---|
149 | </div> |
---|
150 | <div onclick="SwitchMenu('menu_1.3', 'skin/')" id="menu_1.3Title" class="menutitle">HDFS</div> |
---|
151 | <div id="menu_1.3" class="menuitemgroup"> |
---|
152 | <div class="menuitem"> |
---|
153 | <a href="hdfs_user_guide.html">User Guide</a> |
---|
154 | </div> |
---|
155 | <div class="menuitem"> |
---|
156 | <a href="hdfs_design.html">Architecture</a> |
---|
157 | </div> |
---|
158 | <div class="menuitem"> |
---|
159 | <a href="hdfs_shell.html">File System Shell Guide</a> |
---|
160 | </div> |
---|
161 | <div class="menuitem"> |
---|
162 | <a href="hdfs_permissions_guide.html">Permissions Guide</a> |
---|
163 | </div> |
---|
164 | <div class="menuitem"> |
---|
165 | <a href="hdfs_quota_admin_guide.html">Quotas Guide</a> |
---|
166 | </div> |
---|
167 | <div class="menuitem"> |
---|
168 | <a href="SLG_user_guide.html">Synthetic Load Generator Guide</a> |
---|
169 | </div> |
---|
170 | <div class="menuitem"> |
---|
171 | <a href="libhdfs.html">C API libhdfs</a> |
---|
172 | </div> |
---|
173 | </div> |
---|
174 | <div onclick="SwitchMenu('menu_1.4', 'skin/')" id="menu_1.4Title" class="menutitle">HOD</div> |
---|
175 | <div id="menu_1.4" class="menuitemgroup"> |
---|
176 | <div class="menuitem"> |
---|
177 | <a href="hod_user_guide.html">User Guide</a> |
---|
178 | </div> |
---|
179 | <div class="menuitem"> |
---|
180 | <a href="hod_admin_guide.html">Admin Guide</a> |
---|
181 | </div> |
---|
182 | <div class="menuitem"> |
---|
183 | <a href="hod_config_guide.html">Config Guide</a> |
---|
184 | </div> |
---|
185 | </div> |
---|
186 | <div onclick="SwitchMenu('menu_1.5', 'skin/')" id="menu_1.5Title" class="menutitle">Miscellaneous</div> |
---|
187 | <div id="menu_1.5" class="menuitemgroup"> |
---|
188 | <div class="menuitem"> |
---|
189 | <a href="api/index.html">API Docs</a> |
---|
190 | </div> |
---|
191 | <div class="menuitem"> |
---|
192 | <a href="jdiff/changes.html">API Changes</a> |
---|
193 | </div> |
---|
194 | <div class="menuitem"> |
---|
195 | <a href="http://wiki.apache.org/hadoop/">Wiki</a> |
---|
196 | </div> |
---|
197 | <div class="menuitem"> |
---|
198 | <a href="http://wiki.apache.org/hadoop/FAQ">FAQ</a> |
---|
199 | </div> |
---|
200 | <div class="menuitem"> |
---|
201 | <a href="releasenotes.html">Release Notes</a> |
---|
202 | </div> |
---|
203 | <div class="menuitem"> |
---|
204 | <a href="changes.html">Change Log</a> |
---|
205 | </div> |
---|
206 | </div> |
---|
207 | <div id="credit"></div> |
---|
208 | <div id="roundbottom"> |
---|
209 | <img style="display: none" class="corner" height="15" width="15" alt="" src="skin/images/rc-b-l-15-1body-2menu-3menu.png"></div> |
---|
210 | <!--+ |
---|
211 | |alternative credits |
---|
212 | +--> |
---|
213 | <div id="credit2"></div> |
---|
214 | </div> |
---|
215 | <!--+ |
---|
216 | |end Menu |
---|
217 | +--> |
---|
218 | <!--+ |
---|
219 | |start content |
---|
220 | +--> |
---|
221 | <div id="content"> |
---|
222 | <div title="Portable Document Format" class="pdflink"> |
---|
223 | <a class="dida" href="fair_scheduler.pdf"><img alt="PDF -icon" src="skin/images/pdfdoc.gif" class="skin"><br> |
---|
224 | PDF</a> |
---|
225 | </div> |
---|
226 | <h1>Fair Scheduler Guide</h1> |
---|
227 | <div id="minitoc-area"> |
---|
228 | <ul class="minitoc"> |
---|
229 | <li> |
---|
230 | <a href="#Purpose">Purpose</a> |
---|
231 | </li> |
---|
232 | <li> |
---|
233 | <a href="#Introduction">Introduction</a> |
---|
234 | </li> |
---|
235 | <li> |
---|
236 | <a href="#Installation">Installation</a> |
---|
237 | </li> |
---|
238 | <li> |
---|
239 | <a href="#Configuring+the+Fair+scheduler">Configuring the Fair scheduler</a> |
---|
240 | </li> |
---|
241 | <li> |
---|
242 | <a href="#Administration"> Administration</a> |
---|
243 | </li> |
---|
244 | <li> |
---|
245 | <a href="#Implementation">Implementation</a> |
---|
246 | </li> |
---|
247 | </ul> |
---|
248 | </div> |
---|
249 | |
---|
250 | |
---|
251 | <a name="N1000D"></a><a name="Purpose"></a> |
---|
252 | <h2 class="h3">Purpose</h2> |
---|
253 | <div class="section"> |
---|
254 | <p>This document describes the Fair Scheduler, a pluggable |
---|
255 | Map/Reduce scheduler for Hadoop which provides a way to share |
---|
256 | large clusters.</p> |
---|
257 | </div> |
---|
258 | |
---|
259 | |
---|
260 | <a name="N10017"></a><a name="Introduction"></a> |
---|
261 | <h2 class="h3">Introduction</h2> |
---|
262 | <div class="section"> |
---|
263 | <p>Fair scheduling is a method of assigning resources to jobs |
---|
264 | such that all jobs get, on average, an equal share of resources |
---|
265 | over time. When there is a single job running, that job uses the |
---|
266 | entire cluster. When other jobs are submitted, tasks slots that |
---|
267 | free up are assigned to the new jobs, so that each job gets |
---|
268 | roughly the same amount of CPU time. Unlike the default Hadoop |
---|
269 | scheduler, which forms a queue of jobs, this lets short jobs finish |
---|
270 | in reasonable time while not starving long jobs. It is also a |
---|
271 | reasonable way to share a cluster between a number of users. Finally, |
---|
272 | fair sharing can also work with job priorities - the priorities are |
---|
273 | used as weights to determine the fraction of total compute time that |
---|
274 | each job should get. |
---|
275 | </p> |
---|
276 | <p> |
---|
277 | The scheduler actually organizes jobs further into "pools", and |
---|
278 | shares resources fairly between these pools. By default, there is a |
---|
279 | separate pool for each user, so that each user gets the same share |
---|
280 | of the cluster no matter how many jobs they submit. However, it is |
---|
281 | also possible to set a job's pool based on the user's Unix group or |
---|
282 | any other jobconf property, such as the queue name property used by |
---|
283 | <a href="capacity_scheduler.html">Capacity Scheduler</a>. |
---|
284 | Within each pool, fair sharing is used to share capacity between |
---|
285 | the running jobs. Pools can also be given weights to share the |
---|
286 | cluster non-proportionally in the config file. |
---|
287 | </p> |
---|
288 | <p> |
---|
289 | In addition to providing fair sharing, the Fair Scheduler allows |
---|
290 | assigning guaranteed minimum shares to pools, which is useful for |
---|
291 | ensuring that certain users, groups or production applications |
---|
292 | always get sufficient resources. When a pool contains jobs, it gets |
---|
293 | at least its minimum share, but when the pool does not need its full |
---|
294 | guaranteed share, the excess is split between other running jobs. |
---|
295 | This lets the scheduler guarantee capacity for pools while utilizing |
---|
296 | resources efficiently when these pools don't contain jobs. |
---|
297 | </p> |
---|
298 | <p> |
---|
299 | The Fair Scheduler lets all jobs run by default, but it is also |
---|
300 | possible to limit the number of running jobs per user and per pool |
---|
301 | through the config file. This can be useful when a user must submit |
---|
302 | hundreds of jobs at once, or in general to improve performance if |
---|
303 | running too many jobs at once would cause too much intermediate data |
---|
304 | to be created or too much context-switching. Limiting the jobs does |
---|
305 | not cause any subsequently submitted jobs to fail, only to wait in the |
---|
306 | sheduler's queue until some of the user's earlier jobs finish. Jobs to |
---|
307 | run from each user/pool are chosen in order of priority and then |
---|
308 | submit time, as in the default FIFO scheduler in Hadoop. |
---|
309 | </p> |
---|
310 | <p> |
---|
311 | Finally, the fair scheduler provides several extension points where |
---|
312 | the basic functionality can be extended. For example, the weight |
---|
313 | calculation can be modified to give a priority boost to new jobs, |
---|
314 | implementing a "shortest job first" policy which reduces response |
---|
315 | times for interactive jobs even further. |
---|
316 | </p> |
---|
317 | </div> |
---|
318 | |
---|
319 | |
---|
320 | <a name="N10031"></a><a name="Installation"></a> |
---|
321 | <h2 class="h3">Installation</h2> |
---|
322 | <div class="section"> |
---|
323 | <p> |
---|
324 | To run the fair scheduler in your Hadoop installation, you need to put |
---|
325 | it on the CLASSPATH. The easiest way is to copy the |
---|
326 | <em>hadoop-*-fairscheduler.jar</em> from |
---|
327 | <em>HADOOP_HOME/contrib/fairscheduler</em> to <em>HADOOP_HOME/lib</em>. |
---|
328 | Alternatively you can modify <em>HADOOP_CLASSPATH</em> to include this jar, in |
---|
329 | <em>HADOOP_CONF_DIR/hadoop-env.sh</em> |
---|
330 | |
---|
331 | </p> |
---|
332 | <p> |
---|
333 | In order to compile fair scheduler, from sources execute <em> ant |
---|
334 | package</em> in source folder and copy the |
---|
335 | <em>build/contrib/fair-scheduler/hadoop-*-fairscheduler.jar</em> |
---|
336 | to <em>HADOOP_HOME/lib</em> |
---|
337 | |
---|
338 | </p> |
---|
339 | <p> |
---|
340 | You will also need to set the following property in the Hadoop config |
---|
341 | file <em>HADOOP_CONF_DIR/mapred-site.xml</em> to have Hadoop use |
---|
342 | the fair scheduler: <br> |
---|
343 | |
---|
344 | <span class="codefrag"><property></span> |
---|
345 | <br> |
---|
346 | |
---|
347 | <span class="codefrag"> <name>mapred.jobtracker.taskScheduler</name></span> |
---|
348 | <br> |
---|
349 | |
---|
350 | <span class="codefrag"> <value>org.apache.hadoop.mapred.FairScheduler</value></span> |
---|
351 | <br> |
---|
352 | |
---|
353 | <span class="codefrag"></property></span> |
---|
354 | |
---|
355 | </p> |
---|
356 | <p> |
---|
357 | Once you restart the cluster, you can check that the fair scheduler |
---|
358 | is running by going to http://<jobtracker URL>/scheduler |
---|
359 | on the JobTracker's web UI. A "job scheduler administration" page should |
---|
360 | be visible there. This page is described in the Administration section. |
---|
361 | </p> |
---|
362 | </div> |
---|
363 | |
---|
364 | |
---|
365 | <a name="N10070"></a><a name="Configuring+the+Fair+scheduler"></a> |
---|
366 | <h2 class="h3">Configuring the Fair scheduler</h2> |
---|
367 | <div class="section"> |
---|
368 | <p> |
---|
369 | The following properties can be set in mapred-site.xml to configure |
---|
370 | the fair scheduler: |
---|
371 | </p> |
---|
372 | <table class="ForrestTable" cellspacing="1" cellpadding="4"> |
---|
373 | |
---|
374 | <tr> |
---|
375 | |
---|
376 | <th colspan="1" rowspan="1">Name</th><th colspan="1" rowspan="1">Description</th> |
---|
377 | |
---|
378 | </tr> |
---|
379 | |
---|
380 | <tr> |
---|
381 | |
---|
382 | <td colspan="1" rowspan="1"> |
---|
383 | mapred.fairscheduler.allocation.file |
---|
384 | </td> |
---|
385 | <td colspan="1" rowspan="1"> |
---|
386 | Specifies an absolute path to an XML file which contains the |
---|
387 | allocations for each pool, as well as the per-pool and per-user |
---|
388 | limits on number of running jobs. If this property is not |
---|
389 | provided, allocations are not used.<br> |
---|
390 | This file must be in XML format, and can contain three types of |
---|
391 | elements: |
---|
392 | <ul> |
---|
393 | |
---|
394 | <li>pool elements, which may contain elements for minMaps, |
---|
395 | minReduces, maxRunningJobs (limit the number of jobs from the |
---|
396 | pool to run at once),and weight (to share the cluster |
---|
397 | non-proportionally with other pools). |
---|
398 | </li> |
---|
399 | |
---|
400 | <li>user elements, which may contain a maxRunningJobs to limit |
---|
401 | jobs. Note that by default, there is a separate pool for each |
---|
402 | user, so these may not be necessary; they are useful, however, |
---|
403 | if you create a pool per user group or manually assign jobs |
---|
404 | to pools.</li> |
---|
405 | |
---|
406 | <li>A userMaxJobsDefault element, which sets the default running |
---|
407 | job limit for any users whose limit is not specified.</li> |
---|
408 | |
---|
409 | </ul> |
---|
410 | |
---|
411 | <br> |
---|
412 | Example Allocation file is listed below :<br> |
---|
413 | |
---|
414 | <span class="codefrag"><?xml version="1.0"?> </span> |
---|
415 | <br> |
---|
416 | |
---|
417 | <span class="codefrag"><allocations></span> |
---|
418 | <br> |
---|
419 | |
---|
420 | <span class="codefrag"> <pool name="sample_pool"></span> |
---|
421 | <br> |
---|
422 | |
---|
423 | <span class="codefrag"> <minMaps>5</minMaps></span> |
---|
424 | <br> |
---|
425 | |
---|
426 | <span class="codefrag"> <minReduces>5</minReduces></span> |
---|
427 | <br> |
---|
428 | |
---|
429 | <span class="codefrag"> <weight>2.0</weight></span> |
---|
430 | <br> |
---|
431 | |
---|
432 | <span class="codefrag"> </pool></span> |
---|
433 | <br> |
---|
434 | |
---|
435 | <span class="codefrag"> <user name="sample_user"></span> |
---|
436 | <br> |
---|
437 | |
---|
438 | <span class="codefrag"> <maxRunningJobs>6</maxRunningJobs></span> |
---|
439 | <br> |
---|
440 | |
---|
441 | <span class="codefrag"> </user></span> |
---|
442 | <br> |
---|
443 | |
---|
444 | <span class="codefrag"> <userMaxJobsDefault>3</userMaxJobsDefault></span> |
---|
445 | <br> |
---|
446 | |
---|
447 | <span class="codefrag"></allocations></span> |
---|
448 | |
---|
449 | <br> |
---|
450 | This example creates a pool sample_pool with a guarantee of 5 map |
---|
451 | slots and 5 reduce slots. The pool also has a weight of 2.0, meaning |
---|
452 | it has a 2x higher share of the cluster than other pools (the default |
---|
453 | weight is 1). Finally, the example limits the number of running jobs |
---|
454 | per user to 3, except for sample_user, who can run 6 jobs concurrently. |
---|
455 | Any pool not defined in the allocations file will have no guaranteed |
---|
456 | capacity and a weight of 1.0. Also, any pool or user with no max |
---|
457 | running jobs set in the file will be allowed to run an unlimited |
---|
458 | number of jobs. |
---|
459 | </td> |
---|
460 | |
---|
461 | </tr> |
---|
462 | |
---|
463 | <tr> |
---|
464 | |
---|
465 | <td colspan="1" rowspan="1"> |
---|
466 | mapred.fairscheduler.assignmultiple |
---|
467 | </td> |
---|
468 | <td colspan="1" rowspan="1"> |
---|
469 | Allows the scheduler to assign both a map task and a reduce task |
---|
470 | on each heartbeat, which improves cluster throughput when there |
---|
471 | are many small tasks to run. Boolean value, default: false. |
---|
472 | </td> |
---|
473 | |
---|
474 | </tr> |
---|
475 | |
---|
476 | <tr> |
---|
477 | |
---|
478 | <td colspan="1" rowspan="1"> |
---|
479 | mapred.fairscheduler.sizebasedweight |
---|
480 | </td> |
---|
481 | <td colspan="1" rowspan="1"> |
---|
482 | Take into account job sizes in calculating their weights for fair |
---|
483 | sharing.By default, weights are only based on job priorities. |
---|
484 | Setting this flag to true will make them based on the size of the |
---|
485 | job (number of tasks needed) as well,though not linearly |
---|
486 | (the weight will be proportional to the log of the number of tasks |
---|
487 | needed). This lets larger jobs get larger fair shares while still |
---|
488 | providing enough of a share to small jobs to let them finish fast. |
---|
489 | Boolean value, default: false. |
---|
490 | </td> |
---|
491 | |
---|
492 | </tr> |
---|
493 | |
---|
494 | <tr> |
---|
495 | |
---|
496 | <td colspan="1" rowspan="1"> |
---|
497 | mapred.fairscheduler.poolnameproperty |
---|
498 | </td> |
---|
499 | <td colspan="1" rowspan="1"> |
---|
500 | Specify which jobconf property is used to determine the pool that a |
---|
501 | job belongs in. String, default: user.name (i.e. one pool for each |
---|
502 | user). Some other useful values to set this to are: <br> |
---|
503 | |
---|
504 | <ul> |
---|
505 | |
---|
506 | <li> group.name (to create a pool per Unix group).</li> |
---|
507 | |
---|
508 | <li>mapred.job.queue.name (the same property as the queue name in |
---|
509 | <a href="capacity_scheduler.html">Capacity Scheduler</a>).</li> |
---|
510 | |
---|
511 | </ul> |
---|
512 | |
---|
513 | </td> |
---|
514 | |
---|
515 | </tr> |
---|
516 | |
---|
517 | <tr> |
---|
518 | |
---|
519 | <td colspan="1" rowspan="1"> |
---|
520 | mapred.fairscheduler.weightadjuster |
---|
521 | </td> |
---|
522 | <td colspan="1" rowspan="1"> |
---|
523 | An extensibility point that lets you specify a class to adjust the |
---|
524 | weights of running jobs. This class should implement the |
---|
525 | <em>WeightAdjuster</em> interface. There is currently one example |
---|
526 | implementation - <em>NewJobWeightBooster</em>, which increases the |
---|
527 | weight of jobs for the first 5 minutes of their lifetime to let |
---|
528 | short jobs finish faster. To use it, set the weightadjuster |
---|
529 | property to the full class name, |
---|
530 | <span class="codefrag">org.apache.hadoop.mapred.NewJobWeightBooster</span> |
---|
531 | NewJobWeightBooster itself provides two parameters for setting the |
---|
532 | duration and boost factor. <br> |
---|
533 | |
---|
534 | <ol> |
---|
535 | |
---|
536 | <li> |
---|
537 | <em>mapred.newjobweightbooster.factor</em> |
---|
538 | Factor by which new jobs weight should be boosted. Default is 3</li> |
---|
539 | |
---|
540 | <li> |
---|
541 | <em>mapred.newjobweightbooster.duration</em> |
---|
542 | Duration in milliseconds, default 300000 for 5 minutes</li> |
---|
543 | |
---|
544 | </ol> |
---|
545 | |
---|
546 | </td> |
---|
547 | |
---|
548 | </tr> |
---|
549 | |
---|
550 | <tr> |
---|
551 | |
---|
552 | <td colspan="1" rowspan="1"> |
---|
553 | mapred.fairscheduler.loadmanager |
---|
554 | </td> |
---|
555 | <td colspan="1" rowspan="1"> |
---|
556 | An extensibility point that lets you specify a class that determines |
---|
557 | how many maps and reduces can run on a given TaskTracker. This class |
---|
558 | should implement the LoadManager interface. By default the task caps |
---|
559 | in the Hadoop config file are used, but this option could be used to |
---|
560 | make the load based on available memory and CPU utilization for example. |
---|
561 | </td> |
---|
562 | |
---|
563 | </tr> |
---|
564 | |
---|
565 | <tr> |
---|
566 | |
---|
567 | <td colspan="1" rowspan="1"> |
---|
568 | mapred.fairscheduler.taskselector: |
---|
569 | </td> |
---|
570 | <td colspan="1" rowspan="1"> |
---|
571 | An extensibility point that lets you specify a class that determines |
---|
572 | which task from within a job to launch on a given tracker. This can be |
---|
573 | used to change either the locality policy (e.g. keep some jobs within |
---|
574 | a particular rack) or the speculative execution algorithm (select |
---|
575 | when to launch speculative tasks). The default implementation uses |
---|
576 | Hadoop's default algorithms from JobInProgress. |
---|
577 | </td> |
---|
578 | |
---|
579 | </tr> |
---|
580 | |
---|
581 | </table> |
---|
582 | </div> |
---|
583 | |
---|
584 | <a name="N10151"></a><a name="Administration"></a> |
---|
585 | <h2 class="h3"> Administration</h2> |
---|
586 | <div class="section"> |
---|
587 | <p> |
---|
588 | The fair scheduler provides support for administration at runtime |
---|
589 | through two mechanisms: |
---|
590 | </p> |
---|
591 | <ol> |
---|
592 | |
---|
593 | <li> |
---|
594 | It is possible to modify pools' allocations |
---|
595 | and user and pool running job limits at runtime by editing the allocation |
---|
596 | config file. The scheduler will reload this file 10-15 seconds after it |
---|
597 | sees that it was modified. |
---|
598 | </li> |
---|
599 | |
---|
600 | <li> |
---|
601 | Current jobs, pools, and fair shares can be examined through the |
---|
602 | JobTracker's web interface, at http://<jobtracker URL>/scheduler. |
---|
603 | On this interface, it is also possible to modify jobs' priorities or |
---|
604 | move jobs from one pool to another and see the effects on the fair |
---|
605 | shares (this requires JavaScript). |
---|
606 | </li> |
---|
607 | |
---|
608 | </ol> |
---|
609 | <p> |
---|
610 | The following fields can be seen for each job on the web interface: |
---|
611 | </p> |
---|
612 | <ul> |
---|
613 | |
---|
614 | <li> |
---|
615 | <em>Submitted</em> - Date and time job was submitted.</li> |
---|
616 | |
---|
617 | <li> |
---|
618 | <em>JobID, User, Name</em> - Job identifiers as on the standard |
---|
619 | web UI.</li> |
---|
620 | |
---|
621 | <li> |
---|
622 | <em>Pool</em> - Current pool of job. Select another value to move job to |
---|
623 | another pool.</li> |
---|
624 | |
---|
625 | <li> |
---|
626 | <em>Priority</em> - Current priority. Select another value to change the |
---|
627 | job's priority</li> |
---|
628 | |
---|
629 | <li> |
---|
630 | <em>Maps/Reduces Finished</em>: Number of tasks finished / total tasks.</li> |
---|
631 | |
---|
632 | <li> |
---|
633 | <em>Maps/Reduces Running</em>: Tasks currently running.</li> |
---|
634 | |
---|
635 | <li> |
---|
636 | <em>Map/Reduce Fair Share</em>: The average number of task slots that this |
---|
637 | job should have at any given time according to fair sharing. The actual |
---|
638 | number of tasks will go up and down depending on how much compute time |
---|
639 | the job has had, but on average it will get its fair share amount.</li> |
---|
640 | |
---|
641 | </ul> |
---|
642 | <p> |
---|
643 | In addition, it is possible to turn on an "advanced" view for the web UI, |
---|
644 | by going to http://<jobtracker URL>/scheduler?advanced. This view shows |
---|
645 | four more columns used for calculations internally: |
---|
646 | </p> |
---|
647 | <ul> |
---|
648 | |
---|
649 | <li> |
---|
650 | <em>Maps/Reduce Weight</em>: Weight of the job in the fair sharing |
---|
651 | calculations. This depends on priority and potentially also on |
---|
652 | job size and job age if the <em>sizebasedweight</em> and |
---|
653 | <em>NewJobWeightBooster</em> are enabled.</li> |
---|
654 | |
---|
655 | <li> |
---|
656 | <em>Map/Reduce Deficit</em>: The job's scheduling deficit in machine- |
---|
657 | seconds - the amount of resources it should have gotten according to |
---|
658 | its fair share, minus how many it actually got. Positive deficit means |
---|
659 | the job will be scheduled again in the near future because it needs to |
---|
660 | catch up to its fair share. The scheduler schedules jobs with higher |
---|
661 | deficit ahead of others. Please see the Implementation section of |
---|
662 | this document for details.</li> |
---|
663 | |
---|
664 | </ul> |
---|
665 | </div> |
---|
666 | |
---|
667 | <a name="N101A3"></a><a name="Implementation"></a> |
---|
668 | <h2 class="h3">Implementation</h2> |
---|
669 | <div class="section"> |
---|
670 | <p>There are two aspects to implementing fair scheduling: Calculating |
---|
671 | each job's fair share, and choosing which job to run when a task slot |
---|
672 | becomes available.</p> |
---|
673 | <p>To select jobs to run, the scheduler then keeps track of a |
---|
674 | "deficit" for each job - the difference between the amount of |
---|
675 | compute time it should have gotten on an ideal scheduler, and the amount |
---|
676 | of compute time it actually got. This is a measure of how |
---|
677 | "unfair" we've been to the job. Every few hundred |
---|
678 | milliseconds, the scheduler updates the deficit of each job by looking |
---|
679 | at how many tasks each job had running during this interval vs. its |
---|
680 | fair share. Whenever a task slot becomes available, it is assigned to |
---|
681 | the job with the highest deficit. There is one exception - if there |
---|
682 | were one or more jobs who were not meeting their pool capacity |
---|
683 | guarantees, we only choose among these "needy" jobs (based |
---|
684 | again on their deficit), to ensure that the scheduler meets pool |
---|
685 | guarantees as soon as possible.</p> |
---|
686 | <p> |
---|
687 | The fair shares are calculated by dividing the capacity of the cluster |
---|
688 | among runnable jobs according to a "weight" for each job. By |
---|
689 | default the weight is based on priority, with each level of priority |
---|
690 | having 2x higher weight than the next (for example, VERY_HIGH has 4x the |
---|
691 | weight of NORMAL). However, weights can also be based on job sizes and ages, |
---|
692 | as described in the Configuring section. For jobs that are in a pool, |
---|
693 | fair shares also take into account the minimum guarantee for that pool. |
---|
694 | This capacity is divided among the jobs in that pool according again to |
---|
695 | their weights. |
---|
696 | </p> |
---|
697 | <p>Finally, when limits on a user's running jobs or a pool's running jobs |
---|
698 | are in place, we choose which jobs get to run by sorting all jobs in order |
---|
699 | of priority and then submit time, as in the standard Hadoop scheduler. Any |
---|
700 | jobs that fall after the user/pool's limit in this ordering are queued up |
---|
701 | and wait idle until they can be run. During this time, they are ignored |
---|
702 | from the fair sharing calculations and do not gain or lose deficit (their |
---|
703 | fair share is set to zero).</p> |
---|
704 | </div> |
---|
705 | |
---|
706 | </div> |
---|
707 | <!--+ |
---|
708 | |end content |
---|
709 | +--> |
---|
710 | <div class="clearboth"> </div> |
---|
711 | </div> |
---|
712 | <div id="footer"> |
---|
713 | <!--+ |
---|
714 | |start bottomstrip |
---|
715 | +--> |
---|
716 | <div class="lastmodified"> |
---|
717 | <script type="text/javascript"><!-- |
---|
718 | document.write("Last Published: " + document.lastModified); |
---|
719 | // --></script> |
---|
720 | </div> |
---|
721 | <div class="copyright"> |
---|
722 | Copyright © |
---|
723 | 2008 <a href="http://www.apache.org/licenses/">The Apache Software Foundation.</a> |
---|
724 | </div> |
---|
725 | <!--+ |
---|
726 | |end bottomstrip |
---|
727 | +--> |
---|
728 | </div> |
---|
729 | </body> |
---|
730 | </html> |
---|