-
Notifications
You must be signed in to change notification settings - Fork 39
/
index.html
219 lines (173 loc) · 7.24 KB
/
index.html
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
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
<!DOCTYPE html PUBLIC "-//W3C//DTD HTML 4.01 Transitional//EN">
<html>
<head>
<link rel="StyleSheet" href="style.css" type="text/css">
<title>6.824 Lab 5: Persistence</title>
</head>
<body>
<div align="center">
<h2><a href="../index.html">6.824</a> - Spring 2015</h2>
</div>
<div align="center">
<h1>6.824 Lab 5: Persistence</h1>
</div>
<div align="center">
<h3>Due: May 8 11:59pm</h3>
</div>
<hr>
<h3>Introduction</h3>
<p>
In this lab you'll add persistence to your key/value server. The
overall goal is to be able to recover after the crash and restart of
one or more key/value servers. It's this capability that makes
fault-tolerance really valuable! The specific properties you'll need
to ensure are:
<ul>
<li>If a key/value server crashes (halts cleanly with disk intact),
and is re-started, it should re-join its replica group. The effect on
availability of one or more such crashed key/value servers should be
no worse than if the same servers had been temporarily disconnected
from the network rather than crashing. This ability to re-start
requires that each replica
save its key/value database, Paxos state, and any other needed state
to disk where it can find it after the re-start.
<li>If a key/value server crashes (halts cleanly) and loses its disk
contents, and is re-started, it should acquire a key/value database
and other needed state from the other replicas and re-join its
replica group. If a majority of a replica group simultaneously loses
disk contents, the group cannot ever continue. If a minority simultaneously
lose their disk content, and re-start, the group must recover so that
it can tolerate future crashes.
</ul>
<p>
You do not need to design a high-performance format for the on-disk
data. It is sufficient for a server to store each key/value pair in a
separate file, and to use a few more files to store its other state.
<p>
You do not need to add persistence to the shardmaster. The tester
uses your existing <tt>shardmaster</tt> package.
<p>
This lab requires more thought than you might think.
<p>
You may find
<a href="http://research.google.com/archive/paxos_made_live.html">Paxos
Made Live</a> useful, particularly Section 5.1. Harp may also be worth
reviewing, since it pays special attention to recovery from various
crash and disk-loss failures.
<p>
You should either do Lab 5, or a <a href="../project.html">project</a>, but not both.
<h3>Collaboration Policy</h3>
You must write all the code you hand in for 6.824, except for code
that we give you as part of the assignment. You are not allowed to
look at anyone else's solution, and you are not allowed to look at
code from previous years. You may discuss the assignments with other
students, but you may not look at or copy each others' code. Please do
not publish your code or make it available to future 6.824 students --
for example, please do not make your code visible on github.
<h3>Software</h3>
<p>
Do a <tt>git pull</tt> to get the latest lab software. We supply you
with new skeleton code and new tests in <tt>src/diskv</tt>.
<pre>
$ add 6.824
$ cd ~/6.824
$ git pull
...
$ cd src/diskv
$
</pre>
<h3>Getting Started</h3>
<p>
First merge a copy of your Lab 4 code into <tt>diskv/server.go</tt>,
<tt>common.go</tt>, and <tt>client.go</tt>. Be careful when merging
StartServer(), since it's a bit different from Lab 4. And don't
copy <tt>test_test.go</tt>; Lab 5 has a new set of tests.
<p>
There are a few differences between the Lab 4 and Lab 5 frameworks.
First, StartServer() takes an extra <tt>dir</tt> argument
that tells a key/value server the directory in which it should store
its state (key/value pairs, Paxos state, etc.). A server should only
use files under that directory; it should not use any other files. The
tests give a server the same directory name each time the tests
re-start a given server. StartServer() can tell if it has been re-started
(as opposed to started for the first time) by looking at its
<tt>restart</tt> argument. The tests give each server a different
directory.
<p>
The second big framework difference is that the Lab 5 tests run each
key/value server as a separate UNIX process, rather than as a set of
threads in a single process. <tt>main/diskvd.go</tt> is the <tt>main()</tt>
routine for the key/value server process. The tester runs
<tt>diskvd.go</tt> as a separate program, and <tt>diskvd.go</tt> calls
StartServer().
<p>
After merging your Lab 4 code into <tt>diskv</tt>, you should be able
to pass the tests that print (lab4). These are copies of Lab 4 tests.
<h3>Hints</h3>
<p>
If a server crashes, loses its disk, and re-starts, a potential
problem is that it could participate in Paxos instances that it had
participated in before crashing. Since the server has lost its Paxos
state, it won't participate correctly in these instances. So you must
find a way to ensure that servers that re-join after disk loss only
participate in new instances.
<p>
<tt>diskv/server.go</tt> includes some functions that may be helpful
to you when reading and writing files containing key/value data.
<p>
You may want to use Go's
<a href="http://golang.org/pkg/encoding/gob/">gob</a> package to
format and parse saved state.
Here's an <a href="lab-5-gob.go">example</a>.
As with RPC, if you want to encode structs with gob,
you must capitalize the field names.
<p>
The Lab 5 tester will kill key/value servers so that they stop
executing at a random place, which was not the case in previous labs.
One consequence is that, if your server is writing a file, the tester
might kill it midway through writing the file (much as a real crash
might occur while writing a file). A good way to cause replacement of
a whole file to nevertheless be atomic is to write to a temporary file
in the same directory, and then
call <tt>os.Rename(tempname,realname)</tt>.
<p>
You'll probably have to modify your Paxos implementation, at least so
that it's possible to save and restore a key/value server's Paxos
state.
<p>
Don't run multiple instances of the Lab 5 tests at the same time on
the same machine. They will remove each others' files.
<h3>Handin procedure</h3>
<p>Submit your code via the class's submission website, located here:
<p><a href="https://6824.scripts.mit.edu:444/submit/handin.py/">https://6824.scripts.mit.edu:444/submit/handin.py/</a>
<p>You may use your MIT Certificate or request an API key via email to
log in for the first time.
Your API key (XXX) is displayed once you logged in, which can
be used to upload lab4 from the console as follows.
<pre>
$ cd ~/6.824
$ echo XXX > api.key
$ make lab5
</pre>
You can check the submission website to check if your submission is successful.
<p>You will receive full credit if your software passes
the <tt>test_test.go</tt> tests when we run your software on our
machines. We will use the timestamp of your <strong>last</strong>
submission for the purpose of calculating late days.
<hr>
<address>
Please post questions on <a href="http://piazza.com">Piazza</a>.
<p>
</address>
</body>
</html>
<!-- LocalWords: Paxos Sharded shard sharding sharded Put's src shardmaster
-->
<!-- LocalWords: shardkv cd TestBasic TestUnreliable Go's RPCs RPC's GID px
-->
<!-- LocalWords: kvpaxos Config ErrWrongGroup Handin gzipped czvf whoami tgz
-->
<!-- LocalWords: else's github diskv StartServer dir diskvd os API
-->
<!-- LocalWords: tempname realname api timestamp
-->