summaryrefslogtreecommitdiff
path: root/storage/bdb/test/reputils.tcl
blob: 340e359f26d296719d8f065b0d2a4590866fe719 (plain)
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
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
# See the file LICENSE for redistribution information.
#
# Copyright (c) 2001-2002
#	Sleepycat Software.  All rights reserved.
#
# $Id: reputils.tcl,v 11.34 2002/08/12 17:54:18 sandstro Exp $
#
# Replication testing utilities

# Environment handle for the env containing the replication "communications
# structure" (really a CDB environment).

# The test environment consists of a queue and a # directory (environment)
# per replication site.  The queue is used to hold messages destined for a
# particular site and the directory will contain the environment for the
# site.  So the environment looks like:
#				$testdir
#			 ___________|______________________________
#			/           |              \               \
#		MSGQUEUEDIR	MASTERDIR	CLIENTDIR.0 ...	CLIENTDIR.N-1
#		| | ... |
#		1 2 .. N+1
#
# The master is site 1 in the MSGQUEUEDIR and clients 1-N map to message
# queues 2 - N+1.
#
# The globals repenv(1-N) contain the environment handles for the sites
# with a given id (i.e., repenv(1) is the master's environment.

global queueenv

# Array of DB handles, one per machine ID, for the databases that contain
# messages.
global queuedbs
global machids

global elect_timeout
set elect_timeout 50000000
set drop 0

# Create the directory structure for replication testing.
# Open the master and client environments; store these in the global repenv
# Return the master's environment: "-env masterenv"
#
proc repl_envsetup { envargs largs tnum {nclients 1} {droppct 0} { oob 0 } } {
	source ./include.tcl
	global clientdir
	global drop drop_msg
	global masterdir
	global repenv
	global testdir

	env_cleanup $testdir

	replsetup $testdir/MSGQUEUEDIR

	set masterdir $testdir/MASTERDIR
	file mkdir $masterdir
	if { $droppct != 0 } {
		set drop 1
		set drop_msg [expr 100 / $droppct]
	} else {
		set drop 0
	}

	for { set i 0 } { $i < $nclients } { incr i } {
		set clientdir($i) $testdir/CLIENTDIR.$i
		file mkdir $clientdir($i)
	}

	# Open a master.
	repladd 1
	#
	# Set log smaller than default to force changing files,
	# but big enough so that the tests that use binary files
	# as keys/data can run.
	#
	set lmax [expr 3 * 1024 * 1024]
	set masterenv [eval {berkdb_env -create -log_max $lmax} $envargs \
	    {-home $masterdir -txn -rep_master -rep_transport \
	    [list 1 replsend]}]
	error_check_good master_env [is_valid_env $masterenv] TRUE
	set repenv(master) $masterenv

	# Open clients
	for { set i 0 } { $i < $nclients } { incr i } {
		set envid [expr $i + 2]
		repladd $envid
		set clientenv [eval {berkdb_env -create} $envargs -txn \
		    {-cachesize { 0 10000000 0 }} -lock_max 10000 \
		    {-home $clientdir($i) -rep_client -rep_transport \
		    [list $envid replsend]}]
		error_check_good client_env [is_valid_env $clientenv] TRUE
		set repenv($i) $clientenv
	}
	set repenv($i) NULL
	append largs " -env $masterenv "

	# Process startup messages
	repl_envprocq $tnum $nclients $oob

	return $largs
}

# Process all incoming messages.  Iterate until there are no messages left
# in anyone's queue so that we capture all message exchanges. We verify that
# the requested number of clients matches the number of client environments
# we have.  The oob parameter indicates if we should process the queue
# with out-of-order delivery.  The replprocess procedure actually does
# the real work of processing the queue -- this routine simply iterates
# over the various queues and does the initial setup.

proc repl_envprocq { tnum { nclients 1 } { oob 0 }} {
	global repenv
	global drop

	set masterenv $repenv(master)
	for { set i 0 } { 1 } { incr i } {
		if { $repenv($i) == "NULL"} {
			break
		}
	}
	error_check_good i_nclients $nclients $i

	set name [format "Repl%03d" $tnum]
	berkdb debug_check
	puts -nonewline "\t$name: Processing master/$i client queues"
	set rand_skip 0
	if { $oob } {
		puts " out-of-order"
	} else {
		puts " in order"
	}
	set do_check 1
	set droprestore $drop
	while { 1 } {
		set nproced 0

		if { $oob } {
			set rand_skip [berkdb random_int 2 10]
		}
		incr nproced [replprocessqueue $masterenv 1 $rand_skip]
		for { set i 0 } { $i < $nclients } { incr i } {
			set envid [expr $i + 2]
			if { $oob } {
				set rand_skip [berkdb random_int 2 10]
			}
			set n [replprocessqueue $repenv($i) \
			    $envid $rand_skip]
			incr nproced $n
		}

		if { $nproced == 0 } {
			# Now that we delay requesting records until
			# we've had a few records go by, we should always
			# see that the number of requests is lower than the
			# number of messages that were enqueued.
			for { set i 0 } { $i < $nclients } { incr i } {
				set clientenv $repenv($i)
				set stats [$clientenv rep_stat]
				set queued [getstats $stats  \
				   {Total log records queued}]
				error_check_bad queued_stats \
				    $queued -1
				set requested [getstats $stats \
				   {Log records requested}]
				error_check_bad requested_stats \
				    $requested -1
				if { $queued != 0 && $do_check != 0 } {
					error_check_good num_requested \
					    [expr $requested < $queued] 1
				}

				$clientenv rep_request 1 1
			}

			# If we were dropping messages, we might need
			# to flush the log so that we get everything
			# and end up in the right state.
			if { $drop != 0 } {
				set drop 0
				set do_check 0
				$masterenv rep_flush
				berkdb debug_check
				puts "\t$name: Flushing Master"
			} else {
				break
			}
		}
	}

	# Reset the clients back to the default state in case we
	# have more processing to do.
	for { set i 0 } { $i < $nclients } { incr i } {
		set clientenv $repenv($i)
		$clientenv rep_request 4 128
	}
	set drop $droprestore
}

# Verify that the directories in the master are exactly replicated in
# each of the client environments.

proc repl_envver0 { tnum method { nclients 1 } } {
	global clientdir
	global masterdir
	global repenv

	# Verify the database in the client dir.
	# First dump the master.
	set t1 $masterdir/t1
	set t2 $masterdir/t2
	set t3 $masterdir/t3
	set omethod [convert_method $method]
	set name [format "Repl%03d" $tnum]

	#
	# We are interested in the keys of whatever databases are present
	# in the master environment, so we just call a no-op check function
	# since we have no idea what the contents of this database really is.
	# We just need to walk the master and the clients and make sure they
	# have the same contents.
	#
	set cwd [pwd]
	cd $masterdir
	set stat [catch {glob test*.db} dbs]
	cd $cwd
	if { $stat == 1 } {
		return
	}
	foreach testfile $dbs {
		open_and_dump_file $testfile $repenv(master) $masterdir/t2 \
		    repl_noop dump_file_direction "-first" "-next"

		if { [string compare [convert_method $method] -recno] != 0 } {
			filesort $t2 $t3
			file rename -force $t3 $t2
		}
		for { set i 0 } { $i < $nclients } { incr i } {
			puts "\t$name: Verifying client $i database \
			    $testfile contents."
			open_and_dump_file $testfile $repenv($i) \
			    $t1 repl_noop dump_file_direction "-first" "-next"

			if { [string compare $omethod "-recno"] != 0 } {
				filesort $t1 $t3
			} else {
				catch {file copy -force $t1 $t3} ret
			}
			error_check_good diff_files($t2,$t3) [filecmp $t2 $t3] 0
		}
	}
}

# Remove all the elements from the master and verify that these
# deletions properly propagated to the clients.

proc repl_verdel { tnum method { nclients 1 } } {
	global clientdir
	global masterdir
	global repenv

	# Delete all items in the master.
	set name [format "Repl%03d" $tnum]
	set cwd [pwd]
	cd $masterdir
	set stat [catch {glob test*.db} dbs]
	cd $cwd
	if { $stat == 1 } {
		return
	}
	foreach testfile $dbs {
		puts "\t$name: Deleting all items from the master."
		set txn [$repenv(master) txn]
		error_check_good txn_begin [is_valid_txn $txn \
		    $repenv(master)] TRUE
		set db [berkdb_open -txn $txn -env $repenv(master) $testfile]
		error_check_good reopen_master [is_valid_db $db] TRUE
		set dbc [$db cursor -txn $txn]
		error_check_good reopen_master_cursor \
		    [is_valid_cursor $dbc $db] TRUE
		for { set dbt [$dbc get -first] } { [llength $dbt] > 0 } \
		    { set dbt [$dbc get -next] } {
			error_check_good del_item [$dbc del] 0
		}
		error_check_good dbc_close [$dbc close] 0
		error_check_good txn_commit [$txn commit] 0
		error_check_good db_close [$db close] 0

		repl_envprocq $tnum $nclients

		# Check clients.
		for { set i 0 } { $i < $nclients } { incr i } {
			puts "\t$name: Verifying emptiness of client database $i."

			set db [berkdb_open -env $repenv($i) $testfile]
			error_check_good reopen_client($i) \
			    [is_valid_db $db] TRUE
			set dbc [$db cursor]
			error_check_good reopen_client_cursor($i) \
			    [is_valid_cursor $dbc $db] TRUE

			error_check_good client($i)_empty \
			    [llength [$dbc get -first]] 0

			error_check_good dbc_close [$dbc close] 0
			error_check_good db_close [$db close] 0
		}
	}
}

# Replication "check" function for the dump procs that expect to
# be able to verify the keys and data.
proc repl_noop { k d } {
	return
}

# Close all the master and client environments in a replication test directory.
proc repl_envclose { tnum envargs } {
	source ./include.tcl
	global clientdir
	global encrypt
	global masterdir
	global repenv
	global testdir

	if { [lsearch $envargs "-encrypta*"] !=-1 } {
		set encrypt 1
	}

	# In order to make sure that we have fully-synced and ready-to-verify
	# databases on all the clients, do a checkpoint on the master and
	# process messages in order to flush all the clients.
	set drop 0
	set do_check 0
	set name [format "Repl%03d" $tnum]
	berkdb debug_check
	puts "\t$name: Checkpointing master."
	error_check_good masterenv_ckp [$repenv(master) txn_checkpoint] 0

	# Count clients.
	for { set ncli 0 } { 1 } { incr ncli } {
		if { $repenv($ncli) == "NULL" } {
			break
		}
	}
	repl_envprocq $tnum $ncli

	error_check_good masterenv_close [$repenv(master) close] 0
	verify_dir $masterdir "\t$name: " 0 0 1
	for { set i 0 } { $i < $ncli } { incr i } {
		error_check_good client($i)_close [$repenv($i) close] 0
		verify_dir $clientdir($i) "\t$name: " 0 0 1
	}
	replclose $testdir/MSGQUEUEDIR

}

# Close up a replication group
proc replclose { queuedir } {
	global queueenv queuedbs machids

	foreach m $machids {
		set db $queuedbs($m)
		error_check_good dbr_close [$db close] 0
	}
	error_check_good qenv_close [$queueenv close] 0
	set machids {}
}

# Create a replication group for testing.
proc replsetup { queuedir } {
	global queueenv queuedbs machids

	file mkdir $queuedir
	set queueenv \
	    [berkdb_env -create -txn -lock_max 20000 -home $queuedir]
	error_check_good queueenv [is_valid_env $queueenv] TRUE

	if { [info exists queuedbs] } {
		unset queuedbs
	}
	set machids {}

	return $queueenv
}

# Send function for replication.
proc replsend { control rec fromid toid } {
	global queuedbs queueenv machids
	global drop drop_msg

	#
	# If we are testing with dropped messages, then we drop every
	# $drop_msg time.  If we do that just return 0 and don't do
	# anything.
	#
	if { $drop != 0 } {
		incr drop
		if { $drop == $drop_msg } {
			set drop 1
			return 0
		}
	}
	# XXX
	# -1 is DB_BROADCAST_MID
	if { $toid == -1 } {
		set machlist $machids
	} else {
		if { [info exists queuedbs($toid)] != 1 } {
			error "replsend: machid $toid not found"
		}
		set machlist [list $toid]
	}

	foreach m $machlist {
		# XXX should a broadcast include to "self"?
		if { $m == $fromid } {
			continue
		}

		set db $queuedbs($m)
		set txn [$queueenv txn]
		$db put -txn $txn -append [list $control $rec $fromid]
		error_check_good replsend_commit [$txn commit] 0
	}

	return 0
}

# Nuke all the pending messages for a particular site.
proc replclear { machid } {
	global queuedbs queueenv

	if { [info exists queuedbs($machid)] != 1 } {
		error "FAIL: replclear: machid $machid not found"
	}

	set db $queuedbs($machid)
	set txn [$queueenv txn]
	set dbc [$db cursor -txn $txn]
	for { set dbt [$dbc get -rmw -first] } { [llength $dbt] > 0 } \
	    { set dbt [$dbc get -rmw -next] } {
		error_check_good replclear($machid)_del [$dbc del] 0
	}
	error_check_good replclear($machid)_dbc_close [$dbc close] 0
	error_check_good replclear($machid)_txn_commit [$txn commit] 0
}

# Add a machine to a replication environment.
proc repladd { machid } {
	global queueenv queuedbs machids

	if { [info exists queuedbs($machid)] == 1 } {
		error "FAIL: repladd: machid $machid already exists"
	}

	set queuedbs($machid) [berkdb open -auto_commit \
	    -env $queueenv -create -recno -renumber repqueue$machid.db]
	error_check_good repqueue_create [is_valid_db $queuedbs($machid)] TRUE

	lappend machids $machid
}

# Process a queue of messages, skipping every "skip_interval" entry.
# We traverse the entire queue, but since we skip some messages, we
# may end up leaving things in the queue, which should get picked up
# on a later run.

proc replprocessqueue { dbenv machid { skip_interval 0 } \
    { hold_electp NONE } { newmasterp NONE } } {
	global queuedbs queueenv errorCode

	# hold_electp is a call-by-reference variable which lets our caller
	# know we need to hold an election.
	if { [string compare $hold_electp NONE] != 0 } {
		upvar $hold_electp hold_elect
	}
	set hold_elect 0

	# newmasterp is the same idea, only returning the ID of a master
	# given in a DB_REP_NEWMASTER return.
	if { [string compare $newmasterp NONE] != 0 } {
		upvar $newmasterp newmaster
	}
	set newmaster 0

	set nproced 0

	set txn [$queueenv txn]
	set dbc [$queuedbs($machid) cursor -txn $txn]

	error_check_good process_dbc($machid) \
	    [is_valid_cursor $dbc $queuedbs($machid)] TRUE

	for { set dbt [$dbc get -first] } \
	    { [llength $dbt] != 0 } \
	    { set dbt [$dbc get -next] } {
		set data [lindex [lindex $dbt 0] 1]

		# If skip_interval is nonzero, we want to process messages
		# out of order.  We do this in a simple but slimy way--
		# continue walking with the cursor without processing the
		# message or deleting it from the queue, but do increment
		# "nproced".  The way this proc is normally used, the
		# precise value of nproced doesn't matter--we just don't
		# assume the queues are empty if it's nonzero.  Thus,
		# if we contrive to make sure it's nonzero, we'll always
		# come back to records we've skipped on a later call
		# to replprocessqueue.  (If there really are no records,
		# we'll never get here.)
		#
		# Skip every skip_interval'th record (and use a remainder other
		# than zero so that we're guaranteed to really process at least
		# one record on every call).
		if { $skip_interval != 0 } {
			if { $nproced % $skip_interval == 1 } {
				incr nproced
				continue
			}
		}

		# We have to play an ugly cursor game here:  we currently
		# hold a lock on the page of messages, but rep_process_message
		# might need to lock the page with a different cursor in
		# order to send a response.  So save our recno, close
		# the cursor, and then reopen and reset the cursor.
		set recno [lindex [lindex $dbt 0] 0]
		error_check_good dbc_process_close [$dbc close] 0
		error_check_good txn_commit [$txn commit] 0
		set ret [catch {$dbenv rep_process_message \
		    [lindex $data 2] [lindex $data 0] [lindex $data 1]} res]
		set txn [$queueenv txn]
		set dbc [$queuedbs($machid) cursor -txn $txn]
		set dbt [$dbc get -set $recno]

		if { $ret != 0 } {
			if { [is_substr $res DB_REP_HOLDELECTION] } {
				set hold_elect 1
			} else {
				error "FAIL:[timestamp]\
				    rep_process_message returned $res"
			}
		}

		incr nproced

		$dbc del

		if { $ret == 0 && $res != 0 } {
			if { [is_substr $res DB_REP_NEWSITE] } {
				# NEWSITE;  do nothing.
			} else {
				set newmaster $res
				# Break as soon as we get a NEWMASTER message;
				# our caller needs to handle it.
				break
			}
		}

		if { $hold_elect == 1 } {
			# Break also on a HOLDELECTION, for the same reason.
			break
		}

	}

	error_check_good dbc_close [$dbc close] 0
	error_check_good txn_commit [$txn commit] 0

	# Return the number of messages processed.
	return $nproced
}

set run_repl_flag "-run_repl"

proc extract_repl_args { args } {
	global run_repl_flag

	for { set arg [lindex $args [set i 0]] } \
	    { [string length $arg] > 0 } \
	    { set arg [lindex $args [incr i]] } {
		if { [string compare $arg $run_repl_flag] == 0 } {
			return [lindex $args [expr $i + 1]]
		}
	}
	return ""
}

proc delete_repl_args { args } {
	global run_repl_flag

	set ret {}

	for { set arg [lindex $args [set i 0]] } \
	    { [string length $arg] > 0 } \
	    { set arg [lindex $args [incr i]] } {
		if { [string compare $arg $run_repl_flag] != 0 } {
			lappend ret $arg
		} else {
			incr i
		}
	}
	return $ret
}

global elect_serial
global elections_in_progress
set elect_serial 0

# Start an election in a sub-process.
proc start_election { qdir envstring nsites pri timeout {err "none"}} {
	source ./include.tcl
	global elect_serial elect_timeout elections_in_progress machids

	incr elect_serial

	set t [open "|$tclsh_path >& $testdir/ELECTION_OUTPUT.$elect_serial" w]

	puts $t "source $test_path/test.tcl"
	puts $t "replsetup $qdir"
	foreach i $machids { puts $t "repladd $i" }
	puts $t "set env_cmd \{$envstring\}"
	puts $t "set dbenv \[eval \$env_cmd -errfile \
	    $testdir/ELECTION_ERRFILE.$elect_serial -errpfx FAIL: \]"
# puts "Start election err $err, env $envstring"
	puts $t "\$dbenv test abort $err"
	puts $t "set res \[catch \{\$dbenv rep_elect $nsites $pri \
	    $elect_timeout\} ret\]"
	if { $err != "none" } {
		puts $t "\$dbenv test abort none"
		puts $t "set res \[catch \{\$dbenv rep_elect $nsites $pri \
		    $elect_timeout\} ret\]"
	}
	flush $t

	set elections_in_progress($elect_serial) $t
	return $elect_serial
}

proc close_election { i } {
	global elections_in_progress
	set t $elections_in_progress($i)
	puts $t "\$dbenv close"
	close $t
	unset elections_in_progress($i)
}

proc cleanup_elections { } {
	global elect_serial elections_in_progress

	for { set i 0 } { $i <= $elect_serial } { incr i } {
		if { [info exists elections_in_progress($i)] != 0 } {
			close_election $i
		}
	}

	set elect_serial 0
}