source: trunk/db/fsfs.conf @ 1572

Last change on this file since 1572 was 1572, checked in by pg, 19 months ago
File size: 10.1 KB
Line 
1### This file controls the configuration of the FSFS filesystem.
2
3[memcached-servers]
4### These options name memcached servers used to cache internal FSFS
5### data.  See http://www.danga.com/memcached/ for more information on
6### memcached.  To use memcached with FSFS, run one or more memcached
7### servers, and specify each of them as an option like so:
8# first-server = 127.0.0.1:11211
9# remote-memcached = mymemcached.corp.example.com:11212
10### The option name is ignored; the value is of the form HOST:PORT.
11### memcached servers can be shared between multiple repositories;
12### however, if you do this, you *must* ensure that repositories have
13### distinct UUIDs and paths, or else cached data from one repository
14### might be used by another accidentally.  Note also that memcached has
15### no authentication for reads or writes, so you must ensure that your
16### memcached servers are only accessible by trusted users.
17
18[caches]
19### When a cache-related error occurs, normally Subversion ignores it
20### and continues, logging an error if the server is appropriately
21### configured (and ignoring it with file:// access).  To make
22### Subversion never ignore cache errors, uncomment this line.
23# fail-stop = true
24
25[rep-sharing]
26### To conserve space, the filesystem can optionally avoid storing
27### duplicate representations.  This comes at a slight cost in
28### performance, as maintaining a database of shared representations can
29### increase commit times.  The space savings are dependent upon the size
30### of the repository, the number of objects it contains and the amount of
31### duplication between them, usually a function of the branching and
32### merging process.
33###
34### The following parameter enables rep-sharing in the repository.  It can
35### be switched on and off at will, but for best space-saving results
36### should be enabled consistently over the life of the repository.
37### 'svnadmin verify' will check the rep-cache regardless of this setting.
38### rep-sharing is enabled by default.
39# enable-rep-sharing = true
40
41[deltification]
42### To conserve space, the filesystem stores data as differences against
43### existing representations.  This comes at a slight cost in performance,
44### as calculating differences can increase commit times.  Reading data
45### will also create higher CPU load and the data will be fragmented.
46### Since deltification tends to save significant amounts of disk space,
47### the overall I/O load can actually be lower.
48###
49### The options in this section allow for tuning the deltification
50### strategy.  Their effects on data size and server performance may vary
51### from one repository to another.  Versions prior to 1.8 will ignore
52### this section.
53###
54### The following parameter enables deltification for directories. It can
55### be switched on and off at will, but for best space-saving results
56### should be enabled consistently over the lifetime of the repository.
57### Repositories containing large directories will benefit greatly.
58### In rarely accessed repositories, the I/O overhead may be significant
59### as caches will most likely be low.
60### directory deltification is enabled by default.
61# enable-dir-deltification = true
62###
63### The following parameter enables deltification for properties on files
64### and directories.  Overall, this is a minor tuning option but can save
65### some disk space if you merge frequently or frequently change node
66### properties.  You should not activate this if rep-sharing has been
67### disabled because this may result in a net increase in repository size.
68### property deltification is enabled by default.
69# enable-props-deltification = true
70###
71### During commit, the server may need to walk the whole change history of
72### of a given node to find a suitable deltification base.  This linear
73### process can impact commit times, svnadmin load and similar operations.
74### This setting limits the depth of the deltification history.  If the
75### threshold has been reached, the node will be stored as fulltext and a
76### new deltification history begins.
77### Note, this is unrelated to svn log.
78### Very large values rarely provide significant additional savings but
79### can impact performance greatly - in particular if directory
80### deltification has been activated.  Very small values may be useful in
81### repositories that are dominated by large, changing binaries.
82### Should be a power of two minus 1.  A value of 0 will effectively
83### disable deltification.
84### For 1.8, the default value is 1023; earlier versions have no limit.
85# max-deltification-walk = 1023
86###
87### The skip-delta scheme used by FSFS tends to repeatably store redundant
88### delta information where a simple delta against the latest version is
89### often smaller.  By default, 1.8+ will therefore use skip deltas only
90### after the linear chain of deltas has grown beyond the threshold
91### specified by this setting.
92### Values up to 64 can result in some reduction in repository size for
93### the cost of quickly increasing I/O and CPU costs. Similarly, smaller
94### numbers can reduce those costs at the cost of more disk space.  For
95### rarely read repositories or those containing larger binaries, this may
96### present a better trade-off.
97### Should be a power of two.  A value of 1 or smaller will cause the
98### exclusive use of skip-deltas (as in pre-1.8).
99### For 1.8, the default value is 16; earlier versions use 1.
100# max-linear-deltification = 16
101###
102### After deltification, we compress the data through zlib to minimize on-
103### disk size.  That can be an expensive and ineffective process.  This
104### setting controls the usage of zlib in future revisions.
105### Revisions with highly compressible data in them may shrink in size
106### if the setting is increased but may take much longer to commit.  The
107### time taken to uncompress that data again is widely independent of the
108### compression level.
109### Compression will be ineffective if the incoming content is already
110### highly compressed.  In that case, disabling the compression entirely
111### will speed up commits as well as reading the data.  Repositories with
112### many small compressible files (source code) but also a high percentage
113### of large incompressible ones (artwork) may benefit from compression
114### levels lowered to e.g. 1.
115### Valid values are 0 to 9 with 9 providing the highest compression ratio
116### and 0 disabling it altogether.
117### The default value is 5.
118# compression-level = 5
119
120[packed-revprops]
121### This parameter controls the size (in kBytes) of packed revprop files.
122### Revprops of consecutive revisions will be concatenated into a single
123### file up to but not exceeding the threshold given here.  However, each
124### pack file may be much smaller and revprops of a single revision may be
125### much larger than the limit set here.  The threshold will be applied
126### before optional compression takes place.
127### Large values will reduce disk space usage at the expense of increased
128### latency and CPU usage reading and changing individual revprops.
129### Values smaller than 4 kByte will not improve latency any further and
130### quickly render revprop packing ineffective.
131### revprop-pack-size is 4 kBytes by default for non-compressed revprop
132### pack files and 16 kBytes when compression has been enabled.
133# revprop-pack-size = 4
134###
135### To save disk space, packed revprop files may be compressed.  Standard
136### revprops tend to allow for very effective compression.  Reading and
137### even more so writing, become significantly more CPU intensive.
138### Compressing packed revprops is disabled by default.
139# compress-packed-revprops = false
140
141[io]
142### Parameters in this section control the data access granularity in
143### format 7 repositories and later.  The defaults should translate into
144### decent performance over a wide range of setups.
145###
146### When a specific piece of information needs to be read from disk,  a
147### data block is being read at once and its contents are being cached.
148### If the repository is being stored on a RAID, the block size should be
149### either 50% or 100% of RAID block size / granularity.  Also, your file
150### system blocks/clusters should be properly aligned and sized.  In that
151### setup, each access will hit only one disk (minimizes I/O load) but
152### uses all the data provided by the disk in a single access.
153### For SSD-based storage systems, slightly lower values around 16 kB
154### may improve latency while still maximizing throughput.  If block-read
155### has not been enabled, this will be capped to 4 kBytes.
156### Can be changed at any time but must be a power of 2.
157### block-size is given in kBytes and with a default of 64 kBytes.
158# block-size = 64
159###
160### The log-to-phys index maps data item numbers to offsets within the
161### rev or pack file.  This index is organized in pages of a fixed maximum
162### capacity.  To access an item, the page table and the respective page
163### must be read.
164### This parameter only affects revisions with thousands of changed paths.
165### If you have several extremely large revisions (~1 mio changes), think
166### about increasing this setting.  Reducing the value will rarely result
167### in a net speedup.
168### This is an expert setting.  Must be a power of 2.
169### l2p-page-size is 8192 entries by default.
170# l2p-page-size = 8192
171###
172### The phys-to-log index maps positions within the rev or pack file to
173### to data items,  i.e. describes what piece of information is being
174### stored at any particular offset.  The index describes the rev file
175### in chunks (pages) and keeps a global list of all those pages.  Large
176### pages mean a shorter page table but a larger per-page description of
177### data items in it.  The latency sweetspot depends on the change size
178### distribution but covers a relatively wide range.
179### If the repository contains very large files,  i.e. individual changes
180### of tens of MB each,  increasing the page size will shorten the index
181### file at the expense of a slightly increased latency in sections with
182### smaller changes.
183### For source code repositories, this should be about 16x the block-size.
184### Must be a power of 2.
185### p2l-page-size is given in kBytes and with a default of 1024 kBytes.
186# p2l-page-size = 1024
Note: See TracBrowser for help on using the repository browser.