1 |
NAME |
2 |
BDB - Asynchronous Berkeley DB access |
3 |
|
4 |
SYNOPSIS |
5 |
use BDB; |
6 |
|
7 |
my $env = db_env_create; |
8 |
|
9 |
mkdir "bdtest", 0700; |
10 |
db_env_open |
11 |
$env, |
12 |
"bdtest", |
13 |
BDB::INIT_LOCK | BDB::INIT_LOG | BDB::INIT_MPOOL |
14 |
| BDB::INIT_TXN | BDB::RECOVER | BDB::USE_ENVIRON | BDB::CREATE, |
15 |
0600; |
16 |
|
17 |
$env->set_flags (BDB::AUTO_COMMIT | BDB::TXN_NOSYNC, 1); |
18 |
|
19 |
my $db = db_create $env; |
20 |
db_open $db, undef, "table", undef, BDB::BTREE, BDB::AUTO_COMMIT | BDB::CREATE |
21 |
| BDB::READ_UNCOMMITTED, 0600; |
22 |
db_put $db, undef, "key", "data", 0, sub { |
23 |
db_del $db, undef, "key"; |
24 |
}; |
25 |
db_sync $db; |
26 |
|
27 |
# automatic result processing with AnyEvent: |
28 |
our $FH; open $FH, "<&=" . BDB::poll_fileno; |
29 |
our $WATCHER = AnyEvent->io (fh => $FH, poll => 'r', cb => \&BDB::poll_cb); |
30 |
|
31 |
# automatic result processing with EV: |
32 |
my $WATCHER = EV::io BDB::poll_fileno, EV::READ, \&BDB::poll_cb; |
33 |
|
34 |
# with Glib: |
35 |
add_watch Glib::IO BDB::poll_fileno, |
36 |
in => sub { BDB::poll_cb; 1 }; |
37 |
|
38 |
# or simply flush manually |
39 |
BDB::flush; |
40 |
|
41 |
DESCRIPTION |
42 |
See the BerkeleyDB documentation |
43 |
(<http://www.oracle.com/technology/documentation/berkeley-db/db/index.ht |
44 |
ml>). The BDB API is very similar to the C API (the translation has been |
45 |
very faithful). |
46 |
|
47 |
See also the example sections in the document below and possibly the eg/ |
48 |
subdirectory of the BDB distribution. Last not least see the IO::AIO |
49 |
documentation, as that module uses almost the same asynchronous request |
50 |
model as this module. |
51 |
|
52 |
I know this is woefully inadequate documentation. Send a patch! |
53 |
|
54 |
REQUEST ANATOMY AND LIFETIME |
55 |
Every request method creates a request. which is a C data structure not |
56 |
directly visible to Perl. |
57 |
|
58 |
During their existance, bdb requests travel through the following |
59 |
states, in order: |
60 |
|
61 |
ready |
62 |
Immediately after a request is created it is put into the ready |
63 |
state, waiting for a thread to execute it. |
64 |
|
65 |
execute |
66 |
A thread has accepted the request for processing and is currently |
67 |
executing it (e.g. blocking in read). |
68 |
|
69 |
pending |
70 |
The request has been executed and is waiting for result processing. |
71 |
|
72 |
While request submission and execution is fully asynchronous, result |
73 |
processing is not and relies on the perl interpreter calling |
74 |
"poll_cb" (or another function with the same effect). |
75 |
|
76 |
result |
77 |
The request results are processed synchronously by "poll_cb". |
78 |
|
79 |
The "poll_cb" function will process all outstanding aio requests by |
80 |
calling their callbacks, freeing memory associated with them and |
81 |
managing any groups they are contained in. |
82 |
|
83 |
done |
84 |
Request has reached the end of its lifetime and holds no resources |
85 |
anymore (except possibly for the Perl object, but its connection to |
86 |
the actual aio request is severed and calling its methods will |
87 |
either do nothing or result in a runtime error). |
88 |
|
89 |
BERKELEYDB FUNCTIONS |
90 |
All of these are functions. The create functions simply return a new |
91 |
object and never block. All the remaining functions all take an optional |
92 |
callback as last argument. If it is missing, then the fucntion will be |
93 |
executed synchronously. |
94 |
|
95 |
BDB functions that cannot block (mostly functions that manipulate |
96 |
settings) are method calls on the relevant objects, so the rule of thumb |
97 |
is: if its a method, its not blocking, if its a function, it takes a |
98 |
callback as last argument. |
99 |
|
100 |
In the following, $int signifies an integer return value, "octetstring" |
101 |
is a "binary string" (i.e. a perl string with no character indices |
102 |
>255), "U32" is an unsigned 32 bit integer, "int" is some integer, "NV" |
103 |
is a floating point value. |
104 |
|
105 |
The "SV *" types are generic perl scalars (for input and output of data |
106 |
values), and the "SV *callback" is the optional callback function to |
107 |
call when the request is completed. |
108 |
|
109 |
The various "DB_ENV" etc. arguments are handles return by |
110 |
"db_env_create", "db_create", "txn_begin" and so on. If they have an |
111 |
appended "_ornull" this means they are optional and you can pass "undef" |
112 |
for them, resulting a NULL pointer on the C level. |
113 |
|
114 |
BDB functions |
115 |
Functions in the BDB namespace, exported by default: |
116 |
|
117 |
$env = db_env_create (U32 env_flags = 0) |
118 |
flags: RPCCLIENT |
119 |
|
120 |
db_env_open (DB_ENV *env, octetstring db_home, U32 open_flags, int mode, SV *callback = &PL_sv_undef) |
121 |
open_flags: INIT_CDB INIT_LOCK INIT_LOG INIT_MPOOL INIT_REP INIT_TXN RECOVER RECOVER_FATAL USE_ENVIRON USE_ENVIRON_ROOT CREATE LOCKDOWN PRIVATE REGISTER SYSTEM_MEM |
122 |
db_env_close (DB_ENV *env, U32 flags = 0, SV *callback = &PL_sv_undef) |
123 |
db_env_txn_checkpoint (DB_ENV *env, U32 kbyte = 0, U32 min = 0, U32 flags = 0, SV *callback = &PL_sv_undef) |
124 |
flags: FORCE |
125 |
db_env_lock_detect (DB_ENV *env, U32 flags = 0, U32 atype = DB_LOCK_DEFAULT, SV *dummy = 0, SV *callback = &PL_sv_undef) |
126 |
atype: LOCK_DEFAULT LOCK_EXPIRE LOCK_MAXLOCKS LOCK_MAXWRITE LOCK_MINLOCKS LOCK_MINWRITE LOCK_OLDEST LOCK_RANDOM LOCK_YOUNGEST |
127 |
db_env_memp_sync (DB_ENV *env, SV *dummy = 0, SV *callback = &PL_sv_undef) |
128 |
db_env_memp_trickle (DB_ENV *env, int percent, SV *dummy = 0, SV *callback = &PL_sv_undef) |
129 |
|
130 |
$db = db_create (DB_ENV *env = 0, U32 flags = 0) |
131 |
flags: XA_CREATE |
132 |
|
133 |
db_open (DB *db, DB_TXN_ornull *txnid, octetstring file, octetstring database, int type, U32 flags, int mode, SV *callback = &PL_sv_undef) |
134 |
flags: AUTO_COMMIT CREATE EXCL MULTIVERSION NOMMAP RDONLY READ_UNCOMMITTED THREAD TRUNCATE |
135 |
db_close (DB *db, U32 flags = 0, SV *callback = &PL_sv_undef) |
136 |
flags: DB_NOSYNC |
137 |
db_compact (DB *db, DB_TXN_ornull *txn = 0, SV *start = 0, SV *stop = 0, SV *unused1 = 0, U32 flags = DB_FREE_SPACE, SV *unused2 = 0, SV *callback = &PL_sv_undef) |
138 |
flags: FREELIST_ONLY FREE_SPACE |
139 |
db_sync (DB *db, U32 flags = 0, SV *callback = &PL_sv_undef) |
140 |
db_key_range (DB *db, DB_TXN_ornull *txn, SV *key, SV *key_range, U32 flags = 0, SV *callback = &PL_sv_undef) |
141 |
db_put (DB *db, DB_TXN_ornull *txn, SV *key, SV *data, U32 flags = 0, SV *callback = &PL_sv_undef) |
142 |
flags: APPEND NODUPDATA NOOVERWRITE |
143 |
db_get (DB *db, DB_TXN_ornull *txn, SV *key, SV *data, U32 flags = 0, SV *callback = &PL_sv_undef) |
144 |
flags: CONSUME CONSUME_WAIT GET_BOTH SET_RECNO MULTIPLE READ_COMMITTED READ_UNCOMMITTED RMW |
145 |
db_pget (DB *db, DB_TXN_ornull *txn, SV *key, SV *pkey, SV *data, U32 flags = 0, SV *callback = &PL_sv_undef) |
146 |
flags: CONSUME CONSUME_WAIT GET_BOTH SET_RECNO MULTIPLE READ_COMMITTED READ_UNCOMMITTED RMW |
147 |
db_del (DB *db, DB_TXN_ornull *txn, SV *key, U32 flags = 0, SV *callback = &PL_sv_undef) |
148 |
db_txn_commit (DB_TXN *txn, U32 flags = 0, SV *callback = &PL_sv_undef) |
149 |
flags: TXN_NOSYNC TXN_SYNC |
150 |
db_txn_abort (DB_TXN *txn, SV *callback = &PL_sv_undef) |
151 |
|
152 |
db_c_close (DBC *dbc, SV *callback = &PL_sv_undef) |
153 |
db_c_count (DBC *dbc, SV *count, U32 flags = 0, SV *callback = &PL_sv_undef) |
154 |
db_c_put (DBC *dbc, SV *key, SV *data, U32 flags = 0, SV *callback = &PL_sv_undef) |
155 |
flags: AFTER BEFORE CURRENT KEYFIRST KEYLAST NODUPDATA |
156 |
db_c_get (DBC *dbc, SV *key, SV *data, U32 flags = 0, SV *callback = &PL_sv_undef) |
157 |
flags: CURRENT FIRST GET_BOTH GET_BOTH_RANGE GET_RECNO JOIN_ITEM LAST NEXT NEXT_DUP NEXT_NODUP PREV PREV_DUP PREV_NODUP SET SET_RANGE SET_RECNO READ_UNCOMMITTED MULTIPLE MULTIPLE_KEY RMW |
158 |
db_c_pget (DBC *dbc, SV *key, SV *pkey, SV *data, U32 flags = 0, SV *callback = &PL_sv_undef) |
159 |
db_c_del (DBC *dbc, U32 flags = 0, SV *callback = &PL_sv_undef) |
160 |
|
161 |
db_sequence_open (DB_SEQUENCE *seq, DB_TXN_ornull *txnid, SV *key, U32 flags = 0, SV *callback = &PL_sv_undef) |
162 |
flags: CREATE EXCL |
163 |
db_sequence_close (DB_SEQUENCE *seq, U32 flags = 0, SV *callback = &PL_sv_undef) |
164 |
db_sequence_get (DB_SEQUENCE *seq, DB_TXN_ornull *txnid, int delta, SV *seq_value, U32 flags = DB_TXN_NOSYNC, SV *callback = &PL_sv_undef) |
165 |
flags: TXN_NOSYNC |
166 |
db_sequence_remove (DB_SEQUENCE *seq, DB_TXN_ornull *txnid = 0, U32 flags = 0, SV *callback = &PL_sv_undef) |
167 |
flags: TXN_NOSYNC |
168 |
|
169 |
db_txn_finish (DB_TXN *txn, U32 flags = 0, SV *callback = &PL_sv_undef) |
170 |
This is not actually a Berkeley DB function but a BDB module extension. |
171 |
The background for this exytension is: It is very annoying to have to |
172 |
check every single BDB function for error returns and provide a codepath |
173 |
out of your transaction. While the BDB module still makes this possible, |
174 |
it contains the following extensions: |
175 |
|
176 |
When a transaction-protected function returns any operating system error |
177 |
(errno > 0), BDB will set the "TXN_DEADLOCK" flag on the transaction. |
178 |
This flag is also set by Berkeley DB functions themselves when an |
179 |
operation fails with LOCK_DEADLOCK, and it causes all further operations |
180 |
on that transaction (including "db_txn_commit") to fail. |
181 |
|
182 |
The "db_txn_finish" request will look at this flag, and, if it is set, |
183 |
will automatically call "db_txn_abort" (setting errno to "LOCK_DEADLOCK" |
184 |
if it isn't set to something else yet). If it isn't set, it will call |
185 |
"db_txn_commit" and return the error normally. |
186 |
|
187 |
How to use this? Easy: just write your transaction normally: |
188 |
|
189 |
my $txn = $db_env->txn_begin; |
190 |
db_get $db, $txn, "key", my $data; |
191 |
db_put $db, $txn, "key", $data + 1 unless $! == BDB::NOTFOUND; |
192 |
db_txn_finish $txn; |
193 |
die "transaction failed" if $!; |
194 |
|
195 |
That is, handle only the expected errors. If something unexpected |
196 |
happens (EIO, LOCK_NOTGRANTED or a deadlock in either db_get or db_put), |
197 |
then the remaining requests (db_put in this case) will simply be skipped |
198 |
(they will fail with LOCK_DEADLOCK) and the transaction will be aborted. |
199 |
|
200 |
You can use the "$txn->failed" method to check wether a transaction has |
201 |
failed in this way and abort further processing (excluding |
202 |
"db_txn_finish"). |
203 |
|
204 |
DB_ENV/database environment methods |
205 |
Methods available on DB_ENV/$env handles: |
206 |
|
207 |
DESTROY (DB_ENV_ornull *env) |
208 |
CODE: |
209 |
if (env) |
210 |
env->close (env, 0); |
211 |
|
212 |
$int = $env->set_data_dir (const char *dir) |
213 |
$int = $env->set_tmp_dir (const char *dir) |
214 |
$int = $env->set_lg_dir (const char *dir) |
215 |
$int = $env->set_shm_key (long shm_key) |
216 |
$int = $env->set_cachesize (U32 gbytes, U32 bytes, int ncache = 0) |
217 |
$int = $env->set_flags (U32 flags, int onoff) |
218 |
$env->set_errfile (FILE *errfile = 0) |
219 |
$env->set_msgfile (FILE *msgfile = 0) |
220 |
$int = $env->set_verbose (U32 which, int onoff = 1) |
221 |
$int = $env->set_encrypt (const char *password, U32 flags = 0) |
222 |
$int = $env->set_timeout (NV timeout_seconds, U32 flags = SET_TXN_TIMEOUT) |
223 |
$int = $env->set_mp_max_openfd (int maxopenfd); |
224 |
$int = $env->set_mp_max_write (int maxwrite, int maxwrite_sleep); |
225 |
$int = $env->set_mp_mmapsize (int mmapsize_mb) |
226 |
$int = $env->set_lk_detect (U32 detect = DB_LOCK_DEFAULT) |
227 |
$int = $env->set_lk_max_lockers (U32 max) |
228 |
$int = $env->set_lk_max_locks (U32 max) |
229 |
$int = $env->set_lk_max_objects (U32 max) |
230 |
$int = $env->set_lg_bsize (U32 max) |
231 |
$int = $env->set_lg_max (U32 max) |
232 |
$int = $env->mutex_set_increment (U32 increment) |
233 |
$int = $env->mutex_set_tas_spins (U32 tas_spins) |
234 |
$int = $env->mutex_set_max (U32 max) |
235 |
$int = $env->mutex_set_align (U32 align) |
236 |
|
237 |
$txn = $env->txn_begin (DB_TXN_ornull *parent = 0, U32 flags = 0) |
238 |
flags: READ_COMMITTED READ_UNCOMMITTED TXN_NOSYNC TXN_NOWAIT TXN_SNAPSHOT TXN_SYNC TXN_WAIT TXN_WRITE_NOSYNC |
239 |
|
240 |
Example: |
241 |
use AnyEvent; |
242 |
use BDB; |
243 |
|
244 |
our $FH; open $FH, "<&=" . BDB::poll_fileno; |
245 |
our $WATCHER = AnyEvent->io (fh => $FH, poll => 'r', cb => \&BDB::poll_cb); |
246 |
|
247 |
BDB::min_parallel 8; |
248 |
|
249 |
my $env = db_env_create; |
250 |
|
251 |
mkdir "bdtest", 0700; |
252 |
db_env_open |
253 |
$env, |
254 |
"bdtest", |
255 |
BDB::INIT_LOCK | BDB::INIT_LOG | BDB::INIT_MPOOL | BDB::INIT_TXN | BDB::RECOVER | BDB::USE_ENVIRON | BDB::CREATE, |
256 |
0600; |
257 |
|
258 |
$env->set_flags (BDB::AUTO_COMMIT | BDB::TXN_NOSYNC, 1); |
259 |
|
260 |
DB/database methods |
261 |
Methods available on DB/$db handles: |
262 |
|
263 |
DESTROY (DB_ornull *db) |
264 |
CODE: |
265 |
if (db) |
266 |
{ |
267 |
SV *env = (SV *)db->app_private; |
268 |
db->close (db, 0); |
269 |
SvREFCNT_dec (env); |
270 |
} |
271 |
|
272 |
$int = $db->set_cachesize (U32 gbytes, U32 bytes, int ncache = 0) |
273 |
$int = $db->set_flags (U32 flags) |
274 |
flags: CHKSUM ENCRYPT TXN_NOT_DURABLE |
275 |
Btree: DUP DUPSORT RECNUM REVSPLITOFF |
276 |
Hash: DUP DUPSORT |
277 |
Queue: INORDER |
278 |
Recno: RENUMBER SNAPSHOT |
279 |
|
280 |
$int = $db->set_encrypt (const char *password, U32 flags) |
281 |
$int = $db->set_lorder (int lorder) |
282 |
$int = $db->set_bt_minkey (U32 minkey) |
283 |
$int = $db->set_re_delim (int delim) |
284 |
$int = $db->set_re_pad (int re_pad) |
285 |
$int = $db->set_re_source (char *source) |
286 |
$int = $db->set_re_len (U32 re_len) |
287 |
$int = $db->set_h_ffactor (U32 h_ffactor) |
288 |
$int = $db->set_h_nelem (U32 h_nelem) |
289 |
$int = $db->set_q_extentsize (U32 extentsize) |
290 |
|
291 |
$dbc = $db->cursor (DB_TXN_ornull *txn = 0, U32 flags = 0) |
292 |
flags: READ_COMMITTED READ_UNCOMMITTED WRITECURSOR TXN_SNAPSHOT |
293 |
$seq = $db->sequence (U32 flags = 0) |
294 |
|
295 |
Example: |
296 |
my $db = db_create $env; |
297 |
db_open $db, undef, "table", undef, BDB::BTREE, BDB::AUTO_COMMIT | BDB::CREATE | BDB::READ_UNCOMMITTED, 0600; |
298 |
|
299 |
for (1..1000) { |
300 |
db_put $db, undef, "key $_", "data $_"; |
301 |
|
302 |
db_key_range $db, undef, "key $_", my $keyrange; |
303 |
my ($lt, $eq, $gt) = @$keyrange; |
304 |
} |
305 |
|
306 |
db_del $db, undef, "key $_" for 1..1000; |
307 |
|
308 |
db_sync $db; |
309 |
|
310 |
DB_TXN/transaction methods |
311 |
Methods available on DB_TXN/$txn handles: |
312 |
|
313 |
DESTROY (DB_TXN_ornull *txn) |
314 |
CODE: |
315 |
if (txn) |
316 |
txn->abort (txn); |
317 |
|
318 |
$int = $txn->set_timeout (NV timeout_seconds, U32 flags = SET_TXN_TIMEOUT) |
319 |
flags: SET_LOCK_TIMEOUT SET_TXN_TIMEOUT |
320 |
|
321 |
$bool = $txn->failed |
322 |
# see db_txn_finish documentation, above |
323 |
|
324 |
DBC/cursor methods |
325 |
Methods available on DBC/$dbc handles: |
326 |
|
327 |
DESTROY (DBC_ornull *dbc) |
328 |
CODE: |
329 |
if (dbc) |
330 |
dbc->c_close (dbc); |
331 |
|
332 |
Example: |
333 |
my $c = $db->cursor; |
334 |
|
335 |
for (;;) { |
336 |
db_c_get $c, my $key, my $data, BDB::NEXT; |
337 |
warn "<$!,$key,$data>"; |
338 |
last if $!; |
339 |
} |
340 |
|
341 |
db_c_close $c; |
342 |
|
343 |
DB_SEQUENCE/sequence methods |
344 |
Methods available on DB_SEQUENCE/$seq handles: |
345 |
|
346 |
DESTROY (DB_SEQUENCE_ornull *seq) |
347 |
CODE: |
348 |
if (seq) |
349 |
seq->close (seq, 0); |
350 |
|
351 |
$int = $seq->initial_value (db_seq_t value) |
352 |
$int = $seq->set_cachesize (U32 size) |
353 |
$int = $seq->set_flags (U32 flags) |
354 |
flags: SEQ_DEC SEQ_INC SEQ_WRAP |
355 |
$int = $seq->set_range (db_seq_t min, db_seq_t max) |
356 |
|
357 |
Example: |
358 |
my $seq = $db->sequence; |
359 |
|
360 |
db_sequence_open $seq, undef, "seq", BDB::CREATE; |
361 |
db_sequence_get $seq, undef, 1, my $value; |
362 |
|
363 |
SUPPORT FUNCTIONS |
364 |
EVENT PROCESSING AND EVENT LOOP INTEGRATION |
365 |
$fileno = BDB::poll_fileno |
366 |
Return the *request result pipe file descriptor*. This filehandle |
367 |
must be polled for reading by some mechanism outside this module |
368 |
(e.g. Event or select, see below or the SYNOPSIS). If the pipe |
369 |
becomes readable you have to call "poll_cb" to check the results. |
370 |
|
371 |
See "poll_cb" for an example. |
372 |
|
373 |
BDB::poll_cb |
374 |
Process some outstanding events on the result pipe. You have to call |
375 |
this regularly. Returns the number of events processed. Returns |
376 |
immediately when no events are outstanding. The amount of events |
377 |
processed depends on the settings of "BDB::max_poll_req" and |
378 |
"BDB::max_poll_time". |
379 |
|
380 |
If not all requests were processed for whatever reason, the |
381 |
filehandle will still be ready when "poll_cb" returns. |
382 |
|
383 |
Example: Install an Event watcher that automatically calls |
384 |
BDB::poll_cb with high priority: |
385 |
|
386 |
Event->io (fd => BDB::poll_fileno, |
387 |
poll => 'r', async => 1, |
388 |
cb => \&BDB::poll_cb); |
389 |
|
390 |
BDB::max_poll_reqs $nreqs |
391 |
BDB::max_poll_time $seconds |
392 |
These set the maximum number of requests (default 0, meaning |
393 |
infinity) that are being processed by "BDB::poll_cb" in one call, |
394 |
respectively the maximum amount of time (default 0, meaning |
395 |
infinity) spent in "BDB::poll_cb" to process requests (more |
396 |
correctly the mininum amount of time "poll_cb" is allowed to use). |
397 |
|
398 |
Setting "max_poll_time" to a non-zero value creates an overhead of |
399 |
one syscall per request processed, which is not normally a problem |
400 |
unless your callbacks are really really fast or your OS is really |
401 |
really slow (I am not mentioning Solaris here). Using |
402 |
"max_poll_reqs" incurs no overhead. |
403 |
|
404 |
Setting these is useful if you want to ensure some level of |
405 |
interactiveness when perl is not fast enough to process all requests |
406 |
in time. |
407 |
|
408 |
For interactive programs, values such as 0.01 to 0.1 should be fine. |
409 |
|
410 |
Example: Install an EV watcher that automatically calls BDB::poll_cb |
411 |
with low priority, to ensure that other parts of the program get the |
412 |
CPU sometimes even under high load. |
413 |
|
414 |
# try not to spend much more than 0.1s in poll_cb |
415 |
BDB::max_poll_time 0.1; |
416 |
|
417 |
my $bdb_poll = EV::io BDB::poll_fileno, EV::READ, \&BDB::poll_cb); |
418 |
|
419 |
BDB::poll_wait |
420 |
If there are any outstanding requests and none of them in the result |
421 |
phase, wait till the result filehandle becomes ready for reading |
422 |
(simply does a "select" on the filehandle. This is useful if you |
423 |
want to synchronously wait for some requests to finish). |
424 |
|
425 |
See "nreqs" for an example. |
426 |
|
427 |
BDB::poll |
428 |
Waits until some requests have been handled. |
429 |
|
430 |
Returns the number of requests processed, but is otherwise strictly |
431 |
equivalent to: |
432 |
|
433 |
BDB::poll_wait, BDB::poll_cb |
434 |
|
435 |
BDB::flush |
436 |
Wait till all outstanding BDB requests have been handled. |
437 |
|
438 |
Strictly equivalent to: |
439 |
|
440 |
BDB::poll_wait, BDB::poll_cb |
441 |
while BDB::nreqs; |
442 |
|
443 |
CONTROLLING THE NUMBER OF THREADS |
444 |
BDB::min_parallel $nthreads |
445 |
Set the minimum number of BDB threads to $nthreads. The current |
446 |
default is 8, which means eight asynchronous operations can execute |
447 |
concurrently at any one time (the number of outstanding requests, |
448 |
however, is unlimited). |
449 |
|
450 |
BDB starts threads only on demand, when an BDB request is queued and |
451 |
no free thread exists. Please note that queueing up a hundred |
452 |
requests can create demand for a hundred threads, even if it turns |
453 |
out that everything is in the cache and could have been processed |
454 |
faster by a single thread. |
455 |
|
456 |
It is recommended to keep the number of threads relatively low, as |
457 |
some Linux kernel versions will scale negatively with the number of |
458 |
threads (higher parallelity => MUCH higher latency). With current |
459 |
Linux 2.6 versions, 4-32 threads should be fine. |
460 |
|
461 |
Under most circumstances you don't need to call this function, as |
462 |
the module selects a default that is suitable for low to moderate |
463 |
load. |
464 |
|
465 |
BDB::max_parallel $nthreads |
466 |
Sets the maximum number of BDB threads to $nthreads. If more than |
467 |
the specified number of threads are currently running, this function |
468 |
kills them. This function blocks until the limit is reached. |
469 |
|
470 |
While $nthreads are zero, aio requests get queued but not executed |
471 |
until the number of threads has been increased again. |
472 |
|
473 |
This module automatically runs "max_parallel 0" at program end, to |
474 |
ensure that all threads are killed and that there are no outstanding |
475 |
requests. |
476 |
|
477 |
Under normal circumstances you don't need to call this function. |
478 |
|
479 |
BDB::max_idle $nthreads |
480 |
Limit the number of threads (default: 4) that are allowed to idle |
481 |
(i.e., threads that did not get a request to process within 10 |
482 |
seconds). That means if a thread becomes idle while $nthreads other |
483 |
threads are also idle, it will free its resources and exit. |
484 |
|
485 |
This is useful when you allow a large number of threads (e.g. 100 or |
486 |
1000) to allow for extremely high load situations, but want to free |
487 |
resources under normal circumstances (1000 threads can easily |
488 |
consume 30MB of RAM). |
489 |
|
490 |
The default is probably ok in most situations, especially if thread |
491 |
creation is fast. If thread creation is very slow on your system you |
492 |
might want to use larger values. |
493 |
|
494 |
$oldmaxreqs = BDB::max_outstanding $maxreqs |
495 |
This is a very bad function to use in interactive programs because |
496 |
it blocks, and a bad way to reduce concurrency because it is |
497 |
inexact: Better use an "aio_group" together with a feed callback. |
498 |
|
499 |
Sets the maximum number of outstanding requests to $nreqs. If you to |
500 |
queue up more than this number of requests, the next call to the |
501 |
"poll_cb" (and "poll_some" and other functions calling "poll_cb") |
502 |
function will block until the limit is no longer exceeded. |
503 |
|
504 |
The default value is very large, so there is no practical limit on |
505 |
the number of outstanding requests. |
506 |
|
507 |
You can still queue as many requests as you want. Therefore, |
508 |
"max_oustsanding" is mainly useful in simple scripts (with low |
509 |
values) or as a stop gap to shield against fatal memory overflow |
510 |
(with large values). |
511 |
|
512 |
BDB::set_sync_prepare $cb |
513 |
Sets a callback that is called whenever a request is created without |
514 |
an explicit callback. It has to return two code references. The |
515 |
first is used as the request callback, and the second is called to |
516 |
wait until the first callback has been called. The default |
517 |
implementation works like this: |
518 |
|
519 |
sub { |
520 |
my $status; |
521 |
( |
522 |
sub { $status = $! }, |
523 |
sub { BDB::poll while !defined $status; $! = $status }, |
524 |
) |
525 |
} |
526 |
|
527 |
STATISTICAL INFORMATION |
528 |
BDB::nreqs |
529 |
Returns the number of requests currently in the ready, execute or |
530 |
pending states (i.e. for which their callback has not been invoked |
531 |
yet). |
532 |
|
533 |
Example: wait till there are no outstanding requests anymore: |
534 |
|
535 |
BDB::poll_wait, BDB::poll_cb |
536 |
while BDB::nreqs; |
537 |
|
538 |
BDB::nready |
539 |
Returns the number of requests currently in the ready state (not yet |
540 |
executed). |
541 |
|
542 |
BDB::npending |
543 |
Returns the number of requests currently in the pending state |
544 |
(executed, but not yet processed by poll_cb). |
545 |
|
546 |
FORK BEHAVIOUR |
547 |
This module should do "the right thing" when the process using it forks: |
548 |
|
549 |
Before the fork, BDB enters a quiescent state where no requests can be |
550 |
added in other threads and no results will be processed. After the fork |
551 |
the parent simply leaves the quiescent state and continues |
552 |
request/result processing, while the child frees the request/result |
553 |
queue (so that the requests started before the fork will only be handled |
554 |
in the parent). Threads will be started on demand until the limit set in |
555 |
the parent process has been reached again. |
556 |
|
557 |
In short: the parent will, after a short pause, continue as if fork had |
558 |
not been called, while the child will act as if BDB has not been used |
559 |
yet. |
560 |
|
561 |
MEMORY USAGE |
562 |
Per-request usage: |
563 |
|
564 |
Each aio request uses - depending on your architecture - around 100-200 |
565 |
bytes of memory. In addition, stat requests need a stat buffer (possibly |
566 |
a few hundred bytes), readdir requires a result buffer and so on. Perl |
567 |
scalars and other data passed into aio requests will also be locked and |
568 |
will consume memory till the request has entered the done state. |
569 |
|
570 |
This is not awfully much, so queuing lots of requests is not usually a |
571 |
problem. |
572 |
|
573 |
Per-thread usage: |
574 |
|
575 |
In the execution phase, some aio requests require more memory for |
576 |
temporary buffers, and each thread requires a stack and other data |
577 |
structures (usually around 16k-128k, depending on the OS). |
578 |
|
579 |
KNOWN BUGS |
580 |
Known bugs will be fixed in the next release, except: |
581 |
|
582 |
If you use a transaction in any request, and the request returns |
583 |
with an operating system error or DB_LOCK_NOTGRANTED, the internal |
584 |
TXN_DEADLOCK flag will be set on the transaction. See C<db_txn_finish>, |
585 |
above. |
586 |
|
587 |
SEE ALSO |
588 |
Coro::BDB, IO::AIO. |
589 |
|
590 |
AUTHOR |
591 |
Marc Lehmann <schmorp@schmorp.de> |
592 |
http://home.schmorp.de/ |
593 |
|