Blob


1 .TH VENTI-FMT 8
2 .SH NAME
3 buildindex,
4 checkarenas,
5 checkindex,
6 conf,
7 fmtarenas,
8 fmtbloom,
9 fmtindex,
10 fmtisect,
11 syncindex \- prepare and maintain a venti server
12 .SH SYNOPSIS
13 .PP
14 .B venti/fmtarenas
15 [
16 .B -4Z
17 ]
18 [
19 .B -a
20 .I arenasize
21 ]
22 [
23 .B -b
24 .I blocksize
25 ]
26 .I name
27 .I file
28 .PP
29 .B venti/fmtisect
30 [
31 .B -1Z
32 ]
33 [
34 .B -b
35 .I blocksize
36 ]
37 .I name
38 .I file
39 .PP
40 .B venti/fmtbloom
41 [
42 .B -n
43 .I nblocks
44 |
45 .B -N
46 .I nhash
47 ]
48 [
49 .B -s
50 .I size
51 ]
52 .I file
53 .PP
54 .B venti/fmtindex
55 [
56 .B -a
57 ]
58 .I venti.conf
59 .PP
60 .B venti/conf
61 [
62 .B -w
63 ]
64 .I partition
65 [
66 .I configfile
67 ]
68 .if t .sp 0.5
69 .PP
70 .B venti/buildindex
71 [
72 .B -bd
73 ] [
74 .B -i
75 .I isect
76 ] ... [
77 .B -M
78 .I imemsize
79 ]
80 .I venti.conf
81 .PP
82 .B venti/checkindex
83 [
84 .B -f
85 ]
86 [
87 .B -B
88 .I blockcachesize
89 ]
90 .I venti.conf
91 .I tmp
92 .PP
93 .B venti/checkarenas
94 [
95 .B -afv
96 ]
97 .I file
98 .SH DESCRIPTION
99 These commands aid in the setup, maintenance, and debugging of
100 venti servers.
101 See
102 .MR venti (7)
103 for an overview of the venti system and
104 .MR venti (8)
105 for an overview of the data structures used by the venti server.
106 .PP
107 Note that the units for the various sizes in the following
108 commands can be specified by appending
109 .LR k ,
110 .LR m ,
111 or
112 .LR g
113 to indicate kilobytes, megabytes, or gigabytes respectively.
114 .SS Formatting
115 To prepare a server for its initial use, the arena partitions and
116 the index sections must be formatted individually, with
117 .I fmtarenas
118 and
119 .IR fmtisect .
120 Then the
121 collection of index sections must be combined into a venti
122 index with
123 .IR fmtindex .
124 .PP
125 .I Fmtarenas
126 formats the given
127 .IR file ,
128 typically a disk partition, into an arena partition.
129 The arenas in the partition are given names of the form
130 .IR name%d ,
131 where
132 .I %d
133 is replaced with a sequential number starting at 0.
134 .PP
135 Options to
136 .I fmtarenas
137 are:
138 .TP
139 .BI -a " arenasize
140 The arenas are of
141 .I arenasize
142 bytes. The default is
143 .BR 512M ,
144 which was selected to provide a balance
145 between the number of arenas and the ability to copy an arena to external
146 media such as recordable CDs and tapes.
147 .TP
148 .BI -b " blocksize
149 The size, in bytes, for read and write operations to the file.
150 The size is recorded in the file, and is used by applications that access the arenas.
151 The default is
152 .BR 8k .
153 .TP
154 .B -4
155 Create a `version 4' arena partition for backwards compatibility with old servers.
156 The default is version 5, used by the current venti server.
157 .TP
158 .B -Z
159 Do not zero the data sections of the arenas.
160 Using this option reduces the formatting time
161 but should only be used when it is known that the file was already zeroed.
162 (Version 4 only; version 5 sections are not and do not need to be zeroed.)
163 .PD
164 .PP
165 .I Fmtisect
166 formats the given
167 .IR file ,
168 typically a disk partition, as a venti index section with the specified
169 .IR name .
170 Each of the index sections in a venti configuration must have a unique name.
171 .PP
172 Options to
173 .I fmtisect
174 are:
175 .TP
176 .BI -b " bucketsize
177 The size of an index bucket, in bytes.
178 All the index sections within a index must have the same bucket size.
179 The default is
180 .BR 8k .
181 .TP
182 .B -1
183 Create a `version 1' index section for backwards compatibility with old servers.
184 The default is version 2, used by the current venti server.
185 .TP
186 .B -Z
187 Do not zero the index.
188 Using this option reduces the formatting time
189 but should only be used when it is known that the file was already zeroed.
190 (Version 1 only; version 2 sections are not and do not need to be zeroed.)
191 .PD
192 .PP
193 .I Fmtbloom
194 formats the given
195 .I file
196 as a Bloom filter
197 (see
198 .MR venti (7) ).
199 The options are:
200 .TF "\fL-s\fI size"
201 .PD
202 .TP
203 .BI -n " nblock \fR| " -N " nhash
204 The number of blocks expected to be indexed by the filter
205 or the number of hash functions to use.
206 If the
207 .B -n
208 option
209 is given, it is used, along with the total size of the filter,
210 to compute an appropriate
211 .IR nhash .
212 .TP
213 .BI -s " size
214 The size of the Bloom filter. The default is the total size of the file.
215 In either case,
216 .I size
217 is rounded down to a power of two.
218 .PD
219 .PP
220 The
221 .I file
222 argument in the commands above can be of the form
223 .IB file : lo - hi
224 to specify a range of the file.
225 .I Lo
226 and
227 .I hi
228 are specified in bytes but can have the usual
229 .BI k ,
230 .BI m ,
231 or
232 .B g
233 suffixes.
234 Either
235 .I lo
236 or
237 .I hi
238 may be omitted.
239 This notation eliminates the need to
240 partition raw disks on non-Plan 9 systems.
241 .PP
242 .I Fmtindex
243 reads the configuration file
244 .I venti.conf
245 and initializes the index sections to form a usable index structure.
246 The arena files and index sections must have previously been formatted
247 using
248 .I fmtarenas
249 and
250 .I fmtisect
251 respectively.
252 .PP
253 The function of a venti index is to map a SHA1 fingerprint to a location
254 in the data section of one of the arenas. The index is composed of
255 blocks, each of which contains the mapping for a fixed range of possible
256 fingerprint values.
257 .I Fmtindex
258 determines the mapping between SHA1 values and the blocks
259 of the collection of index sections. Once this mapping has been determined,
260 it cannot be changed without rebuilding the index.
261 The basic assumption in the current implementation is that the index
262 structure is sufficiently empty that individual blocks of the index will rarely
263 overflow. The total size of the index should be about 2% to 10% of
264 the total size of the arenas, but the exact percentage depends both on the
265 index block size and the compressed size of blocks stored.
266 See the discussion in
267 .MR venti (8)
268 for more.
269 .PP
270 .I Fmtindex
271 also computes a mapping between a linear address space and
272 the data section of the collection of arenas. The
273 .B -a
274 option can be used to add additional arenas to an index.
275 To use this feature,
276 add the new arenas to
277 .I venti.conf
278 after the existing arenas and then run
279 .I fmtindex
280 .BR -a .
281 .PP
282 A copy of the above mappings is stored in the header for each of the index sections.
283 These copies enable
284 .I buildindex
285 to restore a single index section without rebuilding the entire index.
286 .PP
287 To make it easier to bootstrap servers, the configuration
288 file can be stored in otherwise empty space
289 at the beginning of any venti partitions using
290 .IR conf .
291 A partition so branded with a configuration file can
292 be used in place of a configuration file when invoking any
293 of the venti commands.
294 By default,
295 .I conf
296 prints the configuration stored in
297 .IR partition .
298 When invoked with the
299 .B -w
300 flag,
301 .I conf
302 reads a configuration file from
303 .I configfile
304 (or else standard input)
305 and stores it in
306 .IR partition .
307 .SS Checking and Rebuilding
308 .PP
309 .I Buildindex
310 populates the index for the Venti system described in
311 .IR venti.conf .
312 The index must have previously been formatted using
313 .IR fmtindex .
314 This command is typically used to build a new index for a Venti
315 system when the old index becomes too small, or to rebuild
316 an index after media failure.
317 Small errors in an index can usually be fixed with
318 .IR checkindex ,
319 but
320 .I checkindex
321 requires a large temporary workspace and
322 .I buildindex
323 does not.
324 .PP
325 Options to
326 .I buildindex
327 are:
328 .TF "\fL-M\fI imemsize"
329 .PD
330 .TP
331 .B -b
332 Reinitialise the Bloom filter, if any.
333 .TP
334 .B -d
335 `Dumb' mode; run all three passes.
336 .TP
337 .BI -i " isect
338 Only rebuild index section
339 .IR isect ;
340 may be repeated to rebuild multiple sections.
341 The name
342 .L none
343 is special and just reads the arenas.
344 .TP
345 .BI -M " imemsize
346 The amount of memory, in bytes, to use for caching raw disk accesses while running
347 .IR buildindex .
348 (This is not a property of the created index.)
349 The usual suffices apply.
350 The default is 256M.
351 .PD
352 .PP
353 .I Checkindex
354 examines the Venti index described in
355 .IR venti.conf .
356 The program detects various error conditions including:
357 blocks that are not indexed, index entries for blocks that do not exist,
358 and duplicate index entries.
359 If requested, an attempt can be made to fix errors that are found.
360 .PP
361 The
362 .I tmp
363 file, usually a disk partition, must be large enough to store a copy of the index.
364 This temporary space is used to perform a merge sort of index entries
365 generated by reading the arenas.
366 .PP
367 Options to
368 .I checkindex
369 are:
370 .TP
371 .BI -B " blockcachesize
372 The amount of memory, in bytes, to use for caching raw disk accesses while running
373 .IR checkindex .
374 The default is 8k.
375 .TP
376 .B -f
377 Attempt to fix any errors that are found.
378 .PD
379 .PP
380 .I Checkarenas
381 examines the Venti arenas contained in the given
382 .IR file .
383 The program detects various error conditions, and optionally attempts
384 to fix any errors that are found.
385 .PP
386 Options to
387 .I checkarenas
388 are:
389 .TP
390 .B -a
391 For each arena, scan the entire data section.
392 If this option is omitted, only the end section of
393 the arena is examined.
394 .TP
395 .B -f
396 Attempt to fix any errors that are found.
397 .TP
398 .B -v
399 Increase the verbosity of output.
400 .PD
401 .SH SOURCE
402 .B \*9/src/cmd/venti/srv
403 .SH SEE ALSO
404 .MR venti (7) ,
405 .MR venti (8)
406 .SH BUGS
407 .I Buildindex
408 should allow an individual index section to be rebuilt.