pg 0.20.0 → 0.21.0

Sign up to get free protection for your applications and to get access to all the features.
@@ -1,231 +0,0 @@
1
- #!/usr/bin/env ruby
2
- # vim: set noet nosta sw=4 ts=4 :
3
- #
4
- # Get the current WAL segment and offset from a master postgresql
5
- # server, and compare slave servers to see how far behind they
6
- # are in MB. This script should be easily modified for use with
7
- # Nagios/Mon/Monit/Zabbix/whatever, or wrapping it in a display loop,
8
- # and is suitable for both WAL shipping or streaming forms of replication.
9
- #
10
- # Mahlon E. Smith <mahlon@martini.nu>
11
- #
12
- # First argument is the master server, all other arguments are treated
13
- # as slave machines.
14
- #
15
- # db_replication.monitor db-master.example.com ...
16
- #
17
-
18
- begin
19
- require 'ostruct'
20
- require 'optparse'
21
- require 'pathname'
22
- require 'etc'
23
- require 'pg'
24
- require 'pp'
25
-
26
- rescue LoadError # 1.8 support
27
- unless Object.const_defined?( :Gem )
28
- require 'rubygems'
29
- retry
30
- end
31
- raise
32
- end
33
-
34
-
35
- ### A class to encapsulate the PG handles.
36
- ###
37
- class PGMonitor
38
-
39
- VERSION = %q$Id: replication_monitor.rb,v 36ca5b412583 2012/04/17 23:32:25 mahlon $
40
-
41
- # When to consider a slave as 'behind', measured in WAL segments.
42
- # The default WAL segment size is 16, so we'll alert after
43
- # missing two WAL files worth of data.
44
- #
45
- LAG_ALERT = 32
46
-
47
- ### Create a new PGMonitor object.
48
- ###
49
- def initialize( opts, hosts )
50
- @opts = opts
51
- @master = hosts.shift
52
- @slaves = hosts
53
- @current_wal = {}
54
- @failures = []
55
- end
56
-
57
- attr_reader :opts, :current_wal, :master, :slaves, :failures
58
-
59
-
60
- ### Perform the connections and check the lag.
61
- ###
62
- def check
63
- # clear prior failures, get current xlog info
64
- @failures = []
65
- return unless self.get_current_wal
66
-
67
- # check all slaves
68
- self.slaves.each do |slave|
69
- begin
70
- slave_db = PG.connect(
71
- :dbname => self.opts.database,
72
- :host => slave,
73
- :port => self.opts.port,
74
- :user => self.opts.user,
75
- :password => self.opts.pass,
76
- :sslmode => 'prefer'
77
- )
78
-
79
- xlog = slave_db.exec( 'SELECT pg_last_xlog_receive_location()' ).getvalue( 0, 0 )
80
- slave_db.close
81
-
82
- lag_in_megs = ( self.find_lag( xlog ).to_f / 1024 / 1024 ).abs
83
- if lag_in_megs >= LAG_ALERT
84
- failures << { :host => slave,
85
- :error => "%0.2fMB behind the master." % [ lag_in_megs ] }
86
- end
87
- rescue => err
88
- failures << { :host => slave, :error => err.message }
89
- end
90
- end
91
- end
92
-
93
-
94
- #########
95
- protected
96
- #########
97
-
98
- ### Ask the master for the current xlog information, to compare
99
- ### to slaves. Returns true on succcess. On failure, populates
100
- ### the failures array and returns false.
101
- ###
102
- def get_current_wal
103
- master_db = PG.connect(
104
- :dbname => self.opts.database,
105
- :host => self.master,
106
- :port => self.opts.port,
107
- :user => self.opts.user,
108
- :password => self.opts.pass,
109
- :sslmode => 'prefer'
110
- )
111
-
112
- self.current_wal[ :segbytes ] = master_db.exec( 'SHOW wal_segment_size' ).
113
- getvalue( 0, 0 ).sub( /\D+/, '' ).to_i << 20
114
-
115
- current = master_db.exec( 'SELECT pg_current_xlog_location()' ).getvalue( 0, 0 )
116
- self.current_wal[ :segment ], self.current_wal[ :offset ] = current.split( /\// )
117
-
118
- master_db.close
119
- return true
120
-
121
- # If we can't get any of the info from the master, then there is no
122
- # point in a comparison with slaves.
123
- #
124
- rescue => err
125
- self.failures << { :host => self.master,
126
- :error => 'Unable to retrieve required info from the master (%s)' % [ err.message ] }
127
-
128
- return false
129
- end
130
-
131
-
132
- ### Given an +xlog+ position from a slave server, return
133
- ### the number of bytes the slave needs to replay before it
134
- ### is caught up to the master.
135
- ###
136
- def find_lag( xlog )
137
- s_segment, s_offset = xlog.split( /\// )
138
- m_segment = self.current_wal[ :segment ]
139
- m_offset = self.current_wal[ :offset ]
140
- m_segbytes = self.current_wal[ :segbytes ]
141
-
142
- return (( m_segment.hex - s_segment.hex ) * m_segbytes) + ( m_offset.hex - s_offset.hex )
143
- end
144
-
145
- end
146
-
147
-
148
- ### Parse command line arguments. Return a struct of global options.
149
- ###
150
- def parse_args( args )
151
- options = OpenStruct.new
152
- options.database = 'postgres'
153
- options.port = 5432
154
- options.user = Etc.getpwuid( Process.uid ).name
155
- options.sslmode = 'prefer'
156
-
157
- opts = OptionParser.new do |opts|
158
- opts.banner = "Usage: #{$0} [options] <master> <slave> [slave2, slave3...]"
159
-
160
- opts.separator ''
161
- opts.separator 'Connection options:'
162
-
163
- opts.on( '-d', '--database DBNAME',
164
- "specify the database to connect to (default: \"#{options.database}\")" ) do |db|
165
- options.database = db
166
- end
167
-
168
- opts.on( '-h', '--host HOSTNAME', 'database server host' ) do |host|
169
- options.host = host
170
- end
171
-
172
- opts.on( '-p', '--port PORT', Integer,
173
- "database server port (default: \"#{options.port}\")" ) do |port|
174
- options.port = port
175
- end
176
-
177
- opts.on( '-U', '--user NAME',
178
- "database user name (default: \"#{options.user}\")" ) do |user|
179
- options.user = user
180
- end
181
-
182
- opts.on( '-W', 'force password prompt' ) do |pw|
183
- print 'Password: '
184
- begin
185
- system 'stty -echo'
186
- options.pass = $stdin.gets.chomp
187
- ensure
188
- system 'stty echo'
189
- puts
190
- end
191
- end
192
-
193
- opts.separator ''
194
- opts.separator 'Other options:'
195
-
196
- opts.on_tail( '--help', 'show this help, then exit' ) do
197
- $stderr.puts opts
198
- exit
199
- end
200
-
201
- opts.on_tail( '--version', 'output version information, then exit' ) do
202
- puts PGMonitor::VERSION
203
- exit
204
- end
205
- end
206
-
207
- opts.parse!( args )
208
- return options
209
- end
210
-
211
-
212
-
213
- if __FILE__ == $0
214
- opts = parse_args( ARGV )
215
- raise ArgumentError, "At least two PostgreSQL servers are required." if ARGV.length < 2
216
- mon = PGMonitor.new( opts, ARGV )
217
-
218
- mon.check
219
- if mon.failures.empty?
220
- puts "All is well!"
221
- exit 0
222
- else
223
- puts "Database replication delayed or broken."
224
- mon.failures.each do |bad|
225
- puts "%s: %s" % [ bad[ :host ], bad[ :error ] ]
226
- end
227
- exit 1
228
- end
229
- end
230
-
231
-
@@ -1,33 +0,0 @@
1
- #!/usr/bin/env ruby1.9.1
2
-
3
- require 'pg'
4
-
5
- db = PG.connect( :dbname => 'test' )
6
- db.exec "DROP TABLE IF EXISTS test"
7
- db.exec "CREATE TABLE test (a INTEGER, b BYTEA)"
8
-
9
- a = 42
10
- b = [1, 2, 3]
11
- db.exec "INSERT INTO test(a, b) VALUES($1::int, $2::bytea)",
12
- [a, {:value => b.pack('N*'), :format => 1}]
13
-
14
- db.exec( "SELECT a::int, b::bytea FROM test LIMIT 1", [], 1 ) do |res|
15
-
16
- res.nfields.times do |i|
17
- puts "Field %d is: %s, a %s (%s) column from table %p" % [
18
- i,
19
- res.fname( i ),
20
- db.exec( "SELECT format_type($1,$2)", [res.ftype(i), res.fmod(1)] ).getvalue(0,0),
21
- res.fformat( i ).zero? ? "string" : "binary",
22
- res.ftable( i ),
23
- ]
24
- end
25
-
26
- res.each do |row|
27
- puts "a = #{row['a'].inspect}"
28
- puts "a (unpacked) = #{row['a'].unpack('N*').inspect}"
29
- puts "b = #{row['b'].unpack('N*').inspect}"
30
- end
31
- end
32
-
33
-
@@ -1,434 +0,0 @@
1
- #!/usr/bin/env ruby
2
- #
3
- # A script to wrap ssh and rsync for PostgreSQL WAL files shipping.
4
- # Mahlon E. Smith <mahlon@martini.nu>
5
- #
6
- # Based off of Joshua Drake's PITRTools concept, but with some important
7
- # differences:
8
- #
9
- # - Only supports PostgreSQL >= 8.3
10
- # - No support for rsync version < 3
11
- # - Only shipping, no client side sync (too much opportunity for failure,
12
- # and it's easy to get a base backup manually)
13
- # - WAL files are only stored once, regardless of how many
14
- # slaves are configured or not responding, and are removed from
15
- # the master when they are no longer needed.
16
- # - Each slave can have completely distinct settings, instead
17
- # of a single set of options applied to all slaves
18
- # - slave sync can be individually paused from the master
19
- # - can run synchronously, or if you have a lot of slaves, threaded async mode
20
- # - It's ruby, instead of python. :)
21
- #
22
- # wal_shipper is configurable via an external YAML file, and will create
23
- # a template on its first run -- you'll need to modify it! It expects
24
- # a directory structure like so:
25
- #
26
- # postgres/
27
- # data/...
28
- # bin/wal_shipper.rb
29
- # etc/wal_shipper.conf <-- YAML settings!
30
- # wal/
31
- #
32
- # It should be loaded from the PostgreSQL master's postgresql.conf
33
- # as such, after putting it into your postgres user homedir under 'bin':
34
- #
35
- # archive_command = '/path/to/postgres_home/bin/wal_shipper.rb %p'
36
- #
37
- # Passwordless ssh keys need to be set up for the postgres user on all
38
- # participating masters and slaves.
39
- #
40
- # You can use any replay method of your choosing on the slaves.
41
- # Here's a nice example using pg_standby, to be put in data/recovery.conf:
42
- #
43
- # restore_command = 'pg_standby -t /tmp/pgrecovery.done -s5 -w0 -c /path/to/postgres_home/wal_files/ %f %p %r'
44
- #
45
- # Or, here's another simple alternative data/recovery.conf, for using WAL shipping
46
- # alongside streaming replication:
47
- #
48
- # standby_mode = 'on'
49
- # primary_conninfo = 'host=master.example.com port=5432 user=repl password=XXXXXXX'
50
- # restore_command = 'cp /usr/local/pgsql/wal/%f %p'
51
- # trigger_file = '/usr/local/pgsql/pg.become_primary'
52
- # archive_cleanup_command = '/usr/local/bin/pg_archivecleanup /usr/local/pgsql/wal %r'
53
- #
54
- #========================================================================================
55
-
56
-
57
- require 'pathname'
58
- require 'yaml'
59
- require 'fileutils'
60
- require 'ostruct'
61
-
62
-
63
- ### Encapsulate WAL shipping functionality.
64
- ###
65
- module WalShipper
66
-
67
- ### Send messages to the PostgreSQL log files.
68
- ###
69
- def log( msg )
70
- return unless @debug
71
- puts "WAL Shipper: %s" % [ msg ]
72
- end
73
-
74
-
75
- ### An object that represents a single destination from the
76
- ### configuration file.
77
- ###
78
- class Destination < OpenStruct
79
- include WalShipper
80
-
81
- ### Create a new WalShipper::Destination object.
82
- def initialize( dest, debug=false )
83
- @debug = debug
84
- super( dest )
85
- self.validate
86
- end
87
-
88
- #########
89
- protected
90
- #########
91
-
92
-
93
- ### Check for required keys and normalize various keys.
94
- ###
95
- def validate
96
- # Check for required destination keys
97
- %w[ label kind ].each do |key|
98
- if self.send( key.to_sym ).nil?
99
- self.log "Destination %p missing required '%s' key." % [ self, key ]
100
- self.invalid = true
101
- end
102
- end
103
-
104
- # Ensure paths are Pathnames for the 'file' destination type.
105
- self.path = Pathname.new( self.path ) if self.kind == 'file'
106
-
107
- if self.kind == 'rsync-ssh'
108
- self.port ||= 22
109
- self.user = self.user ? "#{self.user}@" : ''
110
- end
111
- end
112
- end # Class Destination
113
-
114
-
115
-
116
- ### Class for creating new Destination objects and determining how to
117
- ### ship WAL files to them.
118
- ###
119
- class Dispatcher
120
- include WalShipper
121
-
122
- ### Create a new Shipper object, given a +conf+ hash and a +wal+ file
123
- ### Pathname object.
124
- ###
125
- def initialize( wal, conf )
126
- # Make the config keys instance variables.
127
- conf.each_pair {|key, val| self.instance_variable_set( "@#{key}", val ) }
128
-
129
- # Spool directory check.
130
- #
131
- @spool = Pathname.new( @spool )
132
- @spool.exist? or raise "The configured spool directory (%s) doesn't exist." % [ @spool ]
133
-
134
- # Stop right away if we have disabled shipping.
135
- #
136
- unless @enabled
137
- self.log "WAL shipping is disabled, queuing segment %s" % [ wal.basename ]
138
- exit 1
139
- end
140
-
141
- # Instantiate Destination objects, creating new spool directories
142
- # for each.
143
- #
144
- @destinations.
145
- collect!{|dest| WalShipper::Destination.new( dest, @debug ) }.
146
- reject {|dest| dest.invalid }.
147
- collect do |dest|
148
- dest.spool = @spool + dest.label
149
- dest.spool.mkdir( 0711 ) unless dest.spool.exist?
150
- dest
151
- end
152
-
153
- # Put the WAL file into the spool for processing!
154
- #
155
- @waldir = @spool + 'wal_segments'
156
- @waldir.mkdir( 0711 ) unless @waldir.exist?
157
-
158
- self.log "Copying %s to %s" % [ wal.basename, @waldir ]
159
- FileUtils::cp wal, @waldir
160
-
161
- # 'wal' now references the copy. The original is managed and auto-expired
162
- # by PostgreSQL when a new checkpoint segment it reached.
163
- @wal = @waldir + wal.basename
164
- end
165
-
166
-
167
- ### Create hardlinks for the WAL file into each of the destination directories
168
- ### for separate queueing and recording of what was shipped successfully.
169
- ###
170
- def link
171
- @destinations.each do |dest|
172
- self.log "Linking %s into %s" % [ @wal.basename, dest.spool.basename ]
173
- FileUtils::ln @wal, dest.spool, :force => true
174
- end
175
- end
176
-
177
-
178
- ### Decide to be synchronous or threaded, and delegate each destination
179
- ### to the proper ship method.
180
- ###
181
- def dispatch
182
- # Synchronous mode.
183
- #
184
- unless @async
185
- self.log "Performing a synchronous dispatch."
186
- @destinations.each {|dest| self.dispatch_dest( dest ) }
187
- return
188
- end
189
-
190
- tg = ThreadGroup.new
191
-
192
- # Async, one thread per destination
193
- #
194
- if @async_max.nil? || @async_max.to_i.zero?
195
- self.log "Performing an asynchronous dispatch: one thread per destination."
196
- @destinations.each do |dest|
197
- t = Thread.new do
198
- Thread.current.abort_on_exception = true
199
- self.dispatch_dest( dest )
200
- end
201
- tg.add( t )
202
- end
203
- tg.list.each {|t| t.join }
204
- return
205
- end
206
-
207
- # Async, one thread per destination, in groups of asynx_max size.
208
- #
209
- self.log "Performing an asynchronous dispatch: one thread per destination, %d at a time." % [ @async_max ]
210
- all_dests = @destinations.dup
211
- dest_chunks = []
212
- until all_dests.empty? do
213
- dest_chunks << all_dests.slice!( 0, @async_max )
214
- end
215
-
216
- dest_chunks.each do |chunk|
217
- chunk.each do |dest|
218
- t = Thread.new do
219
- Thread.current.abort_on_exception = true
220
- self.dispatch_dest( dest )
221
- end
222
- tg.add( t )
223
- end
224
-
225
- tg.list.each {|t| t.join }
226
- end
227
-
228
- return
229
- end
230
-
231
-
232
- ### Remove any WAL segments no longer needed by slaves.
233
- ###
234
- def clean_spool
235
- total = 0
236
- @waldir.children.each do |wal|
237
- if wal.stat.nlink == 1
238
- total += wal.unlink
239
- end
240
- end
241
-
242
- self.log "Removed %d WAL segment%s." % [ total, total == 1 ? '' : 's' ]
243
- end
244
-
245
-
246
-
247
- #########
248
- protected
249
- #########
250
-
251
- ### Send WAL segments to remote +dest+ via rsync+ssh.
252
- ### Passwordless keys between the user running this script (postmaster owner)
253
- ### and remote user need to be set up in advance.
254
- ###
255
- def ship_rsync_ssh( dest )
256
- if dest.host.nil?
257
- self.log "Destination %p missing required 'host' key. WAL is queued." % [ dest.host ]
258
- return
259
- end
260
-
261
- rsync_flags = '-zc'
262
- ssh_string = "%s -o ConnectTimeout=%d -o StrictHostKeyChecking=no -p %d" %
263
- [ @ssh, @ssh_timeout || 10, dest.port ]
264
- src_string = ''
265
- dst_string = "%s%s:%s/" % [ dest.user, dest.host, dest.path ]
266
-
267
- # If there are numerous files in the spool dir, it means there was
268
- # an error transferring to this host in the past. Try and ship all
269
- # WAL segments, instead of just the new one. PostgreSQL on the slave
270
- # side will "do the right thing" as they come in, regardless of
271
- # ordering.
272
- #
273
- if dest.spool.children.length > 1
274
- src_string = dest.spool.to_s + '/'
275
- rsync_flags << 'r'
276
- else
277
- src_string = dest.spool + @wal.basename
278
- end
279
-
280
-
281
- ship_wal_cmd = [
282
- @rsync,
283
- @debug ? (rsync_flags << 'vh') : (rsync_flags << 'q'),
284
- '--remove-source-files',
285
- '-e', ssh_string,
286
- src_string, dst_string
287
- ]
288
-
289
- self.log "Running command '%s'" % [ ship_wal_cmd.join(' ') ]
290
- system *ship_wal_cmd
291
-
292
- # Run external notification program on error, if one is configured.
293
- #
294
- unless $?.success?
295
- self.log "Ack! Error while shipping to %p, WAL is queued." % [ dest.label ]
296
- system @error_cmd, dest.label if @error_cmd
297
- end
298
- end
299
-
300
-
301
- ### Copy WAL segments to remote path as set in +dest+.
302
- ### This is useful for longer term PITR, copying to NFS shares, etc.
303
- ###
304
- def ship_file( dest )
305
- if dest.path.nil?
306
- self.log "Destination %p missing required 'path' key. WAL is queued." % [ dest ]
307
- return
308
- end
309
- dest.path.mkdir( 0711 ) unless dest.path.exist?
310
-
311
- # If there are numerous files in the spool dir, it means there was
312
- # an error transferring to this host in the past. Try and ship all
313
- # WAL segments, instead of just the new one. PostgreSQL on the slave
314
- # side will "do the right thing" as they come in, regardless of
315
- # ordering.
316
- #
317
- if dest.spool.children.length > 1
318
- dest.spool.children.each do |wal|
319
- wal.unlink if self.copy_file( wal, dest.path, dest.label, dest.compress )
320
- end
321
- else
322
- wal = dest.spool + @wal.basename
323
- wal.unlink if self.copy_file( wal, dest.path, dest.label, dest.compress )
324
- end
325
- end
326
-
327
-
328
- ### Given a +wal+ Pathname, a +path+ destination, and the destination
329
- ### label, copy and optionally compress a WAL file.
330
- ###
331
- def copy_file( wal, path, label, compress=false )
332
- dest_file = path + wal.basename
333
- FileUtils::cp wal, dest_file
334
- if compress
335
- system *[ 'gzip', '-f', dest_file ]
336
- raise "Error while compressing: %s" % [ wal.basename ] unless $?.success?
337
- end
338
- self.log "Copied %s%s to %s." %
339
- [ wal.basename, compress ? ' (and compressed)' : '', path ]
340
- return true
341
- rescue => err
342
- self.log "Ack! Error while copying '%s' (%s) to %p, WAL is queued." %
343
- [ wal.basename, err.message, path ]
344
- system @error_cmd, label if @error_cmd
345
- return false
346
- end
347
-
348
-
349
- ### Figure out how to send the WAL file to its intended destination +dest+.
350
- ###
351
- def dispatch_dest( dest )
352
- if ! dest.enabled.nil? && ! dest.enabled
353
- self.log "Skipping explicity disabled destination %p, WAL is queued." % [ dest.label ]
354
- return
355
- end
356
-
357
- # Send to the appropriate method. ( rsync-ssh --> ship_rsync_ssh )
358
- #
359
- meth = ( 'ship_' + dest.kind.gsub(/-/, '_') ).to_sym
360
- if WalShipper::Dispatcher.method_defined?( meth )
361
- self.send( meth, dest )
362
- else
363
- self.log "Unknown destination kind %p for %p. WAL is queued." % [ dest.kind, dest.label ]
364
- end
365
- end
366
- end
367
- end
368
-
369
- # Ship the WAL file!
370
- #
371
- if __FILE__ == $0
372
- CONFIG_DIR = Pathname.new( __FILE__ ).dirname.parent + 'etc'
373
- CONFIG = CONFIG_DIR + 'wal_shipper.conf'
374
-
375
- unless CONFIG.exist?
376
- CONFIG_DIR.mkdir( 0711 ) unless CONFIG_DIR.exist?
377
- CONFIG.open('w') {|conf| conf.print(DATA.read) }
378
- CONFIG.chmod( 0644 )
379
- puts "No WAL shipping configuration found, default file created."
380
- end
381
-
382
- wal = ARGV[0] or raise "No WAL file was specified on the command line."
383
- wal = Pathname.new( wal )
384
- conf = YAML.load( CONFIG.read )
385
-
386
- shipper = WalShipper::Dispatcher.new( wal, conf )
387
- shipper.link
388
- shipper.dispatch
389
- shipper.clean_spool
390
- end
391
-
392
-
393
- __END__
394
- ---
395
- # Spool from pg_xlog to the working area?
396
- # This must be set to 'true' for wal shipping to function!
397
- enabled: false
398
-
399
- # Log everything to the PostgreSQL log files?
400
- debug: true
401
-
402
- # The working area for WAL segments.
403
- spool: /opt/local/var/db/postgresql84/wal
404
-
405
- # With multiple slaves, ship WAL in parallel, or be synchronous?
406
- async: false
407
-
408
- # Put a ceiling on the parallel threads?
409
- # '0' or removing this option uses a thread for each destination,
410
- # regardless of how many you have. Keep in mind that's 16 * destination
411
- # count megs of simultaneous bandwidth.
412
- async_max: 5
413
-
414
- # Paths and settings for various binaries.
415
- rsync: /usr/bin/rsync
416
- ssh: /usr/bin/ssh
417
- ssh_timeout: 10
418
-
419
- destinations:
420
-
421
- - label: rsync-example
422
- port: 2222
423
- kind: rsync-ssh
424
- host: localhost
425
- user: postgres
426
- path: wal # relative to the user's homedir on the remote host
427
- enabled: false
428
-
429
- - label: file-example
430
- kind: file
431
- compress: true
432
- enabled: true
433
- path: /tmp/someplace
434
-