1nbdkit-scan-filter(1) NBDKIT nbdkit-scan-filter(1)
2
6 nbdkit-scan-filter - scan disk prefetching data ahead of sequential
7 reads
8
10 nbdkit --filter=scan PLUGIN [scan-ahead=false] [scan-clock=false]
11 [scan-forever=true] [scan-size=]NN
12 nbdkit --filter=scan --filter=cache PLUGIN
14 nbdkit --filter=scan --filter=cow PLUGIN cow-on-cache=true
16
18 "nbdkit-scan-filter" is a filter that scans the disk prefetching data.
19 It is sometimes useful if you expect that the client will read the disk
20 sequentially.
21 The basic operation of the filter is that when a client connects, the
23 filter will start issuing ".cache" (prefetch) requests to the plugin
24 across the whole disk. Plugins which support this command will
25 prefetch the data, making subsequent reads faster. For plugins which
26 do not support this command, you can inject nbdkit-cache-filter(1)
27 below (after) this filter, giving approximately the same effect.
28 nbdkit-cow-filter(1) can be used instead of nbdkit-cache-filter, if you
29 add the "cow-on-cache=true" option.
30 Various parameters can be used to tune scanning, although the defaults
32 should be suitable in most cases.
33 A similar filter is nbdkit-readahead-filter(1).
35 Limitations
37 In a number of significant cases this filter will do nothing. The
38 filter will print a warning message if this happens.
39 Thread model must be parallel*
41 For example nbdkit-curl-plugin(1) only supports
42 "serialize_requests", and so this filter cannot perform prefetches
43 in parallel with the read requests.
44 Only scans while clients are connected*
46 The current filter only scans while there is at least one client
47 connected.
48 Only scans the default export*
50 The current filter only scans the default export and ignores all
51 clients connecting to the non-default export name.
52 *We may be able to lift these restrictions in future.
54 Underlying filters or plugin must support ".cache" (prefetch)
56 Very many plugins do not have the concept of prefetching and/or do
57 not implement the ".cache" callback, and so there is no way for
58 this filter to issue prefetches.
59 You can usually get around this by adding --filter=cache after this
61 filter as explained above.
62 Prefetching the whole disk may load it all into cache
64 In particular if you use this filter together with
65 nbdkit-cache-filter(1) or nbdkit-cow-filter(1), they will cache the
66 whole content of the plugin into a temporary file. This may be
67 many gigabytes of data, consuming all space in /var/tmp. Of course
68 this is the whole point of using this filter, but you should be
69 aware of it.
70 If using the cache filter, the total size of the cache can be
72 limited (see "CACHE MAXIMUM SIZE" in nbdkit-cache-filter(1)).
73
75 scan-ahead=false
76 By default the filter tries to stay ahead of incoming read
77 requests. That is to say, it starts prefetching at the beginning
78 of the disk and continues incrementally, but if the client issues a
79 read beyond the current prefetch point then the filter skips
80 forward and begins prefetching after the read.
81 However if you set this parameter to false, then this behaviour is
83 disabled. The filter simply prefetches sequentially regardless of
84 client requests.
85 scan-clock=false
87 By default, if all clients disconnect and then another client
88 connects, prefetching resumes at the same place in the disk. (Like
89 stopping and starting a clock.)
90 If you set this parameter to false, then the filter starts
92 prefetching from the beginning of the disk again.
93 scan-forever=true
95 By default the filter scans over the disk once and then stops.
96 If you set this parameter to true, then after the disk has been
98 prefetched completely, the filter goes back to the beginning and
99 starts over, repeating this for as long as nbdkit is running and
100 there are clients connected.
101 scan-size=NN
103 This parameter controls the prefetch block size. The default is
104 "2M". This must be a power of 2 and most plugins will have their
105 own limits on the amount of data they can prefetch in a single
106 request.
107
109 $filterdir/nbdkit-scan-filter.so
110 The filter.
111 Use "nbdkit --dump-config" to find the location of $filterdir.
113
115 "nbdkit-scan-filter" first appeared in nbdkit 1.32.
116
118 nbdkit(1), nbdkit-cache-filter(1), nbdkit-cow-filter(1),
119 nbdkit-file-plugin(1), nbdkit-readahead-filter(1),
120 nbdkit-retry-filter(1), nbdkit-torrent-plugin(1),
121 nbdkit-vddk-plugin(1), nbdkit-filter(3), qemu-img(1).
122
124 Richard W.M. Jones
125
127 Copyright (C) 2019-2022 Red Hat Inc.
128
130 Redistribution and use in source and binary forms, with or without
131 modification, are permitted provided that the following conditions are
132 met:
133 • Redistributions of source code must retain the above copyright
135 notice, this list of conditions and the following disclaimer.
136 • Redistributions in binary form must reproduce the above copyright
138 notice, this list of conditions and the following disclaimer in the
139 documentation and/or other materials provided with the
140 distribution.
141 • Neither the name of Red Hat nor the names of its contributors may
143 be used to endorse or promote products derived from this software
144 without specific prior written permission.
145 THIS SOFTWARE IS PROVIDED BY RED HAT AND CONTRIBUTORS ''AS IS'' AND ANY
147 EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
148 IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
149 PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL RED HAT OR CONTRIBUTORS BE
150 LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
151 CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
152 SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR
153 BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY,
154 WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR
155 OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF
156 ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
157nbdkit-1.32.5 2023-01-03 nbdkit-scan-filter(1)