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
|
=head1 NAME
nbdkit-memory-plugin - nbdkit virtual memory (RAM disk) plugin
=head1 SYNOPSIS
nbdkit memory [size=]SIZE [allocator=sparse|malloc|zstd]
=head1 DESCRIPTION
C<nbdkit-memory-plugin> is a plugin for L<nbdkit(1)> which stores a
single disk image in virtual memory, and discards it when nbdkit
exits. This plugin can be used for testing or where you don't care
about the final content of the disk image.
All nbdkit clients will see the same disk content, initially all
zeroes.
By default the disk image is stored in memory using a sparse array.
The allocated parts of the disk image cannot be larger than physical
RAM plus swap, less whatever is being used by the rest of the system.
Other allocators are available, see L</ALLOCATORS> below. All
allocators store the image in memory. If you want to allocate more
space than this use L<nbdkit-file-plugin(1)> backed by a temporary
file instead.
Using the sparse allocator the virtual size can be as large as you
like, up to the maximum supported by nbdkit (S<2⁶³-1 bytes>). This
limit is tested when nbdkit is compiled, and it should work on all
platforms and architectures supported by nbdkit.
=head1 EXAMPLES
Create a one gigabyte sparse RAM disk:
nbdkit memory 1G
If you want to loop mount the above disk, see L<nbdkit-loop(1)>.
Create the largest possible RAM disk:
nbdkit memory $(( 2**63 - 1 ))
=head1 PARAMETERS
=over 4
=item [B<size=>]SIZE
Specify the virtual size of the disk image.
This parameter is required.
C<size=> __IS_MAGIC__
=item B<allocator=sparse>
=item B<allocator=malloc>[,B<mlock=true>]
=item B<allocator=zstd>
(nbdkit E<ge> 1.22)
Select the backend allocation strategy. See L</ALLOCATORS> below.
The default is sparse.
=back
=head1 NOTES
=head2 Preloading small amounts of data
If you want an in-memory disk image preinitialized with a small amount
of data specified on the command line, look at
L<nbdkit-data-plugin(1)> instead. Note by "small" this does not mean
that the virtual disk image must be small, but that the amount of data
initially stored sparsely is small enough to specify on the command
line.
=head2 Preloading large amounts of data
If you want to preload a large amount of data (eg. a disk image) into
the memory plugin, use L<qemu-img(1)> or L<nbdcopy(1)>:
$ rm -f pid
$ nbdkit -P pid memory 10G
Wait for nbdkit to become ready to accept connections:
$ while [ ! -f pid ]; do sleep 1; done
Preload Fedora disk image using qemu-img:
$ virt-builder fedora-28 --size=10G
$ qemu-img convert -p -n fedora-28.img nbd:localhost:10809
If you have libnbd E<ge> 1.4, you can use L<nbdcopy(1)> as an
alternative:
$ nbdcopy -p fedora-28.img nbd://localhost
=head1 ALLOCATORS
Since nbdkit E<ge> 1.22 several allocation strategies are available
using the C<allocator> parameter.
=over 4
=item B<allocator=sparse>
The disk image is stored in memory using a sparse array. The sparse
array uses a simple two level page table with a fixed page size. The
allocated parts of the disk image cannot be larger than physical RAM
plus swap, less whatever is being used by the rest of the system. The
aim of the sparse array implementation is to support extremely large
images for testing, although it won't necessarily be efficient for
that use case. However it should also be reasonably efficient for
normal disk sizes.
The virtual size of the disk can be as large as you like, up to the
maximum supported by nbdkit (S<2⁶³-1 bytes>).
This is the default, and was the only allocator available before
S<nbdkit 1.22>.
=item B<allocator=malloc>
=item B<allocator=malloc,mlock=true>
The disk image is stored directly in memory allocated using
L<malloc(3)> on the heap. No sparseness is possible: you must have
enough memory for the whole disk. Very large virtual sizes will
usually fail. However this can be faster because the implementation
is simpler and the locking strategy allows more concurrency.
If C<mlock=true> is added then additionally the array is locked into
RAM using L<mlock(2)> (so it should never be swapped out). This
usually requires you to adjust the L<ulimit(1)> associated with the
process and on some operating systems may require you to run nbdkit as
root. (See also the L<nbdkit(1)> I<--swap> option).
The C<mlock=true> feature is only supported on some platforms. Use
S<C<nbdkit memory --dump-plugin>> and check that the output contains
C<mlock=yes>.
=item B<allocator=zstd>
The disk image is stored in a sparse array where each page is
compressed using L<zstd compression|https://facebook.github.io/zstd/>.
Assuming a typical 2:1 compression ratio, this allows you to store
twice as much real data as C<allocator=sparse>, with the trade-off
that the plugin is slightly slower because it has to compress and
decompress each page. Aside from compression, the implementation of
this allocator is similar to C<allocator=sparse>, so in other respects
(such as supporting huge virtual disk sizes) it is the same.
This allocator is only supported if nbdkit was compiled with zstd
support. Use S<C<nbdkit memory --dump-plugin>> and check that the
output contains C<zstd=yes>.
=back
=head1 FILES
=over 4
=item F<$plugindir/nbdkit-memory-plugin.so>
The plugin.
Use C<nbdkit --dump-config> to find the location of C<$plugindir>.
=back
=head1 VERSION
C<nbdkit-memory-plugin> first appeared in nbdkit 1.2.
=head1 SEE ALSO
L<nbdkit(1)>,
L<nbdkit-plugin(3)>,
L<nbdkit-loop(1)>,
L<nbdkit-data-plugin(1)>,
L<nbdkit-file-plugin(1)>,
L<nbdkit-info-plugin(1)>,
L<nbdkit-tmpdisk-plugin(1)>,
L<mlock(2)>,
L<malloc(3)>,
L<qemu-img(1)>,
L<nbdcopy(1)>.
=head1 AUTHORS
Richard W.M. Jones
=head1 COPYRIGHT
Copyright Red Hat
|
