-
Notifications
You must be signed in to change notification settings - Fork 23
/
bup-memtest.html
140 lines (138 loc) · 6.06 KB
/
bup-memtest.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
<!DOCTYPE html PUBLIC "-//W3C//DTD XHTML 1.0 Transitional//EN" "http://www.w3.org/TR/xhtml1/DTD/xhtml1-transitional.dtd">
<html xmlns="http://www.w3.org/1999/xhtml"
><head
><title
>bup-memtest(1) Bup 0.24-83-gdfa9d85</title
><meta http-equiv="Content-Type" content="text/html; charset=UTF-8"
/><meta name="generator" content="pandoc"
/><meta name="author" content="Avery Pennarun <apenwarr@gmail.com>"
/><meta name="date" content="2011-03-01"
/></head
><body
><h1 class="title"
>bup-memtest(1) Bup 0.24-83-gdfa9d85</h1
><h1 id="name"
>NAME</h1
><p
>bup-memtest - test bup memory usage statistics</p
><h1 id="synopsis"
>SYNOPSIS</h1
><p
>bup memtest [options...]</p
><h1 id="description"
>DESCRIPTION</h1
><p
><code
>bup memtest</code
> opens the list of pack indexes in your bup repository, then searches the list for a series of nonexistent objects, printing memory usage statistics after each cycle.</p
><p
>Because of the way Unix systems work, the output will usually show a large (and unchanging) value in the VmSize column, because mapping the index files in the first place takes a certain amount of virtual address space. However, this virtual memory usage is entirely virtual; it doesn't take any of your RAM. Over time, bup uses <em
>parts</em
> of the indexes, which need to be loaded from disk, and this is what causes an increase in the VmRSS column.</p
><h1 id="options"
>OPTIONS</h1
><dl
><dt
>-n, --number=<em
>number</em
></dt
><dd
><p
>set the number of objects to search for during each cycle (ie. before printing a line of output)</p
></dd
><dt
>-c, --cycles=<em
>cycles</em
></dt
><dd
><p
>set the number of cycles (ie. the number of lines of output after the first). The first line of output is always 0 (ie. the baseline before searching for any objects).</p
></dd
><dt
>--ignore-midx</dt
><dd
><p
>ignore any <code
>.midx</code
> files created by <code
>bup midx</code
>. This allows you to compare memory performance with and without using midx.</p
></dd
><dt
>--existing</dt
><dd
><p
>search for existing objects instead of searching for random nonexistent ones. This can greatly affect memory usage and performance. Note that most of the time, <code
>bup save</code
> spends most of its time searching for nonexistent objects, since existing ones are probably in unmodified files that we won't be trying to back up anyway. So the default behaviour reflects real bup performance more accurately. But you might want this option anyway just to make sure you haven't made searching for existing objects much worse than before.</p
></dd
></dl
><h1 id="example"
>EXAMPLE</h1
><pre
><code
>$ bup memtest -n300 -c5
PackIdxList: using 1 index.
VmSize VmRSS VmData VmStk
0 20824 kB 4528 kB 1980 kB 84 kB
300 20828 kB 5828 kB 1984 kB 84 kB
600 20828 kB 6844 kB 1984 kB 84 kB
900 20828 kB 7836 kB 1984 kB 84 kB
1200 20828 kB 8736 kB 1984 kB 84 kB
1500 20828 kB 9452 kB 1984 kB 84 kB
$ bup memtest -n300 -c5 --ignore-midx
PackIdxList: using 361 indexes.
VmSize VmRSS VmData VmStk
0 27444 kB 6552 kB 2516 kB 84 kB
300 27448 kB 15832 kB 2520 kB 84 kB
600 27448 kB 17220 kB 2520 kB 84 kB
900 27448 kB 18012 kB 2520 kB 84 kB
1200 27448 kB 18388 kB 2520 kB 84 kB
1500 27448 kB 18556 kB 2520 kB 84 kB
</code
></pre
><h1 id="discussion"
>DISCUSSION</h1
><p
>When optimizing bup indexing, the first goal is to keep the VmRSS reasonably low. However, it might eventually be necessary to swap in all the indexes, simply because you're searching for a lot of objects, and this will cause your RSS to grow as large as VmSize eventually.</p
><p
>The key word here is <em
>eventually</em
>. As long as VmRSS grows reasonably slowly, the amount of disk activity caused by accessing pack indexes is reasonably small. If it grows quickly, bup will probably spend most of its time swapping index data from disk instead of actually running your backup, so backups will run very slowly.</p
><p
>The purpose of <code
>bup memtest</code
> is to give you an idea of how fast your memory usage is growing, and to help in optimizing bup for better memory use. If you have memory problems you might be asked to send the output of <code
>bup memtest</code
> to help diagnose the problems.</p
><p
>Tip: try using <code
>bup midx -a</code
> or <code
>bup midx -f</code
> to see if it helps reduce your memory usage.</p
><p
>Trivia: index memory usage in bup (or git) is only really a problem when adding a large number of previously unseen objects. This is because for each object, we need to absolutely confirm that it isn't already in the database, which requires us to search through <em
>all</em
> the existing pack indexes to ensure that none of them contain the object in question. In the more obvious case of searching for objects that <em
>do</em
> exist, the objects being searched for are typically related in some way, which means they probably all exist in a small number of packfiles, so memory usage will be constrained to just those packfile indexes.</p
><p
>Since git users typically don't add a lot of files in a single run, git doesn't really need a program like <code
>bup midx</code
>. bup, on the other hand, spends most of its time backing up files it hasn't seen before, so its memory usage patterns are different.</p
><h1 id="see-also"
>SEE ALSO</h1
><p
><code
>bup-midx</code
>(1)</p
><h1 id="bup"
>BUP</h1
><p
>Part of the <code
>bup</code
>(1) suite.</p
></body
></html
>