forked from bloomberg/comdb2
-
Notifications
You must be signed in to change notification settings - Fork 0
/
summarize.c
564 lines (499 loc) · 17 KB
/
summarize.c
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
/*
Copyright 2015 Bloomberg Finance L.P.
Licensed under the Apache License, Version 2.0 (the "License");
you may not use this file except in compliance with the License.
You may obtain a copy of the License at
http://www.apache.org/licenses/LICENSE-2.0
Unless required by applicable law or agreed to in writing, software
distributed under the License is distributed on an "AS IS" BASIS,
WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
See the License for the specific language governing permissions and
limitations under the License.
*/
/* This module takes a table as input and produces a "summary" of it as output.
The summary
has every N'th record in the original table. The intended audience is sql
analyze code.
The idea is that for a large enough table, looking at every N'th record for a
reasonable
value of N will give the relative frequency/selectivity data as the entire
table, and this
method of scanning will be faster than a plain walk of the table. */
#include <errno.h>
#include <fcntl.h>
#include <limits.h>
#include <netinet/in.h>
#include <netinet/tcp.h>
#include <stdarg.h>
#include <stdio.h>
#include <stdlib.h>
#include <string.h>
#include <strings.h>
#include <sys/stat.h>
#include <sys/types.h>
#include <sys/socket.h>
#include <sys/socketvar.h>
#include <sys/uio.h>
#include <unistd.h>
#include <stddef.h>
#include <pthread.h>
#include <sbuf2.h>
#include <fcntl.h>
#include <ctrace.h>
#include <net.h>
#include "bdb_int.h"
#include "bdb_cursor.h"
#include "locks.h"
/* need access to berkeley innards here */
#include <build/db.h>
#include <build/db_int.h>
#include <dbinc/db_page.h>
#include <dbinc/btree.h>
#include "dbinc/db_swap.h"
#include "dbinc/hmac.h"
#include <dbinc/crypto.h>
#include <btree/bt_prefix.h>
#include <assert.h>
#include <arpa/nameser_compat.h>
#ifndef BYTE_ORDER
#error "BYTE_ORDER not defined"
#endif
#include "flibc.h"
#include "logmsg.h"
#include "analyze.h"
extern int get_schema_change_in_progress(const char *func, int line);
static double analyze_headroom = 6;
void analyze_set_headroom(uint64_t headroom)
{
if (headroom < 1 || 100 < headroom) {
logmsg(LOGMSG_ERROR,
"Invalid headroom %" PRIu64
". Needs to be between 0 and 100%%\n",
headroom);
return;
}
analyze_headroom = headroom;
logmsg(LOGMSG_USER, "Analyze headroom set to %f%%\n", analyze_headroom);
}
static inline int check_free_space(const char *path)
{
int rc = has_low_headroom(path, 100 - analyze_headroom, 1);
if (rc) {
return BDBERR_MISC;
}
return BDBERR_NOERROR;
}
#define _64K (64 * 1024)
static DB *dbp_from_meta(DB *dbp, DBMETA *meta)
{
uint32_t magic;
if (FLD_ISSET(meta->metaflags, DBMETA_CHKSUM))
F_SET(dbp, DB_AM_CHKSUM);
if (meta->encrypt_alg)
F_SET(dbp, DB_AM_ENCRYPT);
if ((magic = meta->magic) == DB_BTREEMAGIC)
F_CLR(dbp, DB_AM_SWAP);
else if (flibc_intflip(magic) == DB_BTREEMAGIC)
F_SET(dbp, DB_AM_SWAP);
else {
logmsg(LOGMSG_ERROR, "bad meta magic value at %s: %u\n", __func__, magic);
return NULL;
}
dbp->pgsize = F_ISSET(dbp, DB_AM_SWAP) ? flibc_intflip(meta->pagesize)
: meta->pagesize;
dbp->offset_bias = dbp->pgsize > _64K ? dbp->pgsize / _64K : 1;
dbp->type = DB_BTREE;
return dbp;
}
struct sampler {
DB db; /* our DB handle */
bdb_state_type *bdb_state; /* our bdb_state */
struct temp_table *tmptbl; /* temptable to store sampled pages */
struct temp_cursor *tmpcur; /* cursor on the temptable */
int pos; /* to keep track of the index in the page */
void *data; /* payload of the entry at `pos' */
int len; /* length of the payload */
};
int sampler_first(sampler_t *sampler)
{
struct temp_cursor *tmpcur = sampler->tmpcur;
int unused;
if (bdb_temp_table_first(sampler->bdb_state, tmpcur, &unused) != 0)
return IX_EMPTY;
sampler->pos = 0;
return (sampler_next(sampler) == IX_FND) ? IX_FND : IX_EMPTY;
}
int sampler_last(sampler_t *sampler)
{
logmsg(LOGMSG_FATAL, "%s is not implemented.\n", __func__);
abort();
return 0;
}
int sampler_prev(sampler_t *sampler)
{
logmsg(LOGMSG_FATAL, "%s is not implemented.\n", __func__);
abort();
return 0;
}
int sampler_next(sampler_t *sampler)
{
int rc = IX_PASTEOF;
uint8_t pfxbuf[KEYBUF];
DB *dbp = &sampler->db;
PAGE *page;
db_indx_t *inp;
int unused;
struct temp_cursor *tmpcur = sampler->tmpcur;
int ii, n, minlen, memcmprc;
next_leaf:
page = (PAGE *)bdb_temp_table_data(tmpcur);
inp = P_INP(dbp, page);
ii = sampler->pos;
n = NUM_ENT(page);
#ifndef NDEBUG
uint8_t *max = (uint8_t *)page + dbp->pgsize;
#endif
for (; ii < n; ii += 2) {
if (F_ISSET(dbp, DB_AM_SWAP))
inp[ii] = flibc_shortflip(inp[ii]);
BKEYDATA *data = GET_BKEYDATA(dbp, page, ii);
assert((uint8_t *)data < max);
if (F_ISSET(dbp, DB_AM_SWAP))
data->len = flibc_shortflip(data->len);
db_indx_t len;
ASSIGN_ALIGN(db_indx_t, len, data->len);
assert(((uint8_t *)data + len) < max);
if (bk_decompress(dbp, page, &data, pfxbuf, sizeof(pfxbuf)) != 0) {
logmsg(LOGMSG_ERROR, "\ndecompress failed page:%d ii:%d total:%d\n",
page->pgno, ii, n);
continue;
}
ASSIGN_ALIGN(db_indx_t, len, data->len);
/* Even though we sort pages by their 1st key, out-of-order pages
can still occurr if a leaf we have read splits.
We ensure that strictly sorted samples are returned by keeping
reading from the temptable till the 1st entry on current page
is greater than or equal the last entry on previous page. */
if (ii == 0 && sampler->data != NULL) {
minlen = len > sampler->len ? sampler->len : len;
memcmprc = memcmp(sampler->data, data->data, minlen);
if (memcmprc > 0 || (memcmprc == 0 && sampler->len > len))
break;
}
free(sampler->data);
sampler->data = malloc(len);
memcpy(sampler->data, data->data, len);
sampler->len = len;
sampler->pos = ii + 2;
rc = IX_FND;
break;
}
if (rc != IX_FND) {
if (bdb_temp_table_next(sampler->bdb_state, tmpcur, &unused) != 0)
return IX_PASTEOF;
sampler->pos = 0;
goto next_leaf;
}
return rc;
}
void *sampler_key(sampler_t *sampler)
{
return sampler->data;
}
sampler_t *sampler_init(bdb_state_type *bdb_state, int *bdberr)
{
sampler_t *sampler;
sampler = calloc(1, sizeof(sampler_t));
if (sampler == NULL)
goto err;
sampler->tmptbl = bdb_temp_table_create(bdb_state->parent, bdberr);
if (sampler->tmptbl == NULL)
goto err;
sampler->tmpcur =
bdb_temp_table_cursor(bdb_state->parent, sampler->tmptbl, NULL, bdberr);
if (sampler->tmpcur == NULL)
goto err;
sampler->bdb_state = bdb_state;
return sampler;
err:
if (sampler != NULL) {
if (sampler->tmptbl)
bdb_temp_table_close(bdb_state->parent, sampler->tmptbl, bdberr);
free(sampler);
}
return NULL;
}
int sampler_close(sampler_t *sampler)
{
int unused;
if (sampler == NULL)
return 0;
(void)bdb_temp_table_close(sampler->bdb_state, sampler->tmptbl, &unused);
free(sampler->data);
free(sampler);
return 0;
}
int gbl_debug_sleep_in_summarize = 0;
int bdb_summarize_table(bdb_state_type *bdb_state, int ixnum, int comp_pct,
sampler_t **samplerp, unsigned long long *outrecs,
unsigned long long *cmprecs, int *bdberr)
{
DB_ENV *dbenv = bdb_state->dbenv;
int is_hmac = CRYPTO_ON(dbenv);
uint8_t pfxbuf[KEYBUF];
char tmpname[PATH_MAX];
char tran_tmpname[PATH_MAX];
int rc = 0;
DB dbp_ = {0}, *dbp;
PAGE *page = NULL;
unsigned char metabuf[512];
int pgsz = 0;
sampler_t *sampler = *samplerp;
unsigned long long nrecs = 0;
unsigned long long recs_looked_at = 0;
int fd = -1;
int last, now;
#ifdef POSIX_FADV_SEQUENTIAL
/* Release page cache every FADVISE_THRESH many pages. We could make it
a tunable, but for now, leave it hardcoded. */
off_t nread = 0;
const static size_t FADVISE_THRESH = 1024;
int usedio = bdb_attr_get(bdb_state->attr, BDB_ATTR_DIRECTIO);
#endif
if (comp_pct > 100 || comp_pct < 1) {
*bdberr = BDBERR_BADARGS;
rc = -1;
goto done;
}
if (!bdb_state->parent) {
*bdberr = BDBERR_BADARGS;
rc = -1;
goto done;
}
rc = check_free_space(bdb_state->dir);
if (rc != BDBERR_NOERROR) {
*bdberr = rc;
rc = -1;
goto done;
}
sampler = sampler_init(bdb_state, bdberr);
if (sampler == NULL) {
rc = -1;
goto done;
}
*bdberr = BDBERR_NOERROR;
if (ixnum < 0) {
/* we only support this for indices - no need to summarize data */
*bdberr = BDBERR_BADARGS;
goto done;
}
rc = bdb_get_index_filename(bdb_state, ixnum, tmpname, sizeof(tmpname),
bdberr);
if (rc) {
if (rc == DB_LOCK_DEADLOCK)
rc = BDBERR_DEADLOCK;
goto done;
}
bdb_trans(tmpname, tran_tmpname);
logmsg(LOGMSG_DEBUG, "open %s\n", tran_tmpname);
fd = open(tran_tmpname, O_RDONLY);
if (fd == -1) {
logmsg(LOGMSG_ERROR, "can't open input db???: %d %s\n", errno,
strerror(errno));
rc = -1;
goto done;
}
rc = read(fd, metabuf, sizeof(metabuf));
if (rc != sizeof(metabuf)) {
logmsg(LOGMSG_ERROR, "can't read meta page\n");
goto done;
}
if ((dbp = dbp_from_meta(&dbp_, (DBMETA *)metabuf)) == NULL) {
rc = -1;
goto done;
}
pgsz = dbp->pgsize;
page = malloc(pgsz);
#ifndef NDEBUG
uint8_t *max = (uint8_t *)page + pgsz;
#endif
rc = lseek(fd, 0, SEEK_SET);
if (rc) {
logmsg(LOGMSG_ERROR, "can't rewind to start of file\n");
goto done;
}
#ifdef POSIX_FADV_SEQUENTIAL
// inform kernel that we will be accessing file sequentially
(void)posix_fadvise(fd, 0, 0, POSIX_FADV_SEQUENTIAL);
#endif
last = comdb2_time_epoch();
for (rc = read(fd, page, pgsz); rc == pgsz; rc = read(fd, page, pgsz)) {
#ifdef POSIX_FADV_SEQUENTIAL
/* Periodically hint the OS to release pages we've read. Only do so
when directio is enabled for this operation will likely force out
useful cached pages otherwise. */
if (usedio && ((++nread) % FADVISE_THRESH) == 0)
(void)posix_fadvise(fd, (nread - FADVISE_THRESH) * pgsz, FADVISE_THRESH * pgsz, POSIX_FADV_DONTNEED);
#endif
/* If it is not a leaf page, continue reading the file. */
if (!ISLEAF(page))
continue;
if (gbl_debug_sleep_in_summarize) {
sleep(1);
}
int ret;
uint8_t *chksum = NULL;
/* If we have checksums, use them to verify we don't have
a partial page. If the checksum doesn't match,
just skip the page. This should be rare
(only happen for pagesizes larger than default). */
size_t sumlen = 0;
if (F_ISSET(dbp, DB_AM_CHKSUM)) {
chksum_t algo = IS_CRC32C(page) ? algo_crc32c : algo_hash4;
switch (TYPE(page)) {
case P_HASHMETA:
case P_BTREEMETA:
case P_QAMMETA:
chksum = ((BTMETA *)page)->chksum;
sumlen = DBMETASIZE;
break;
default:
chksum = P_CHKSUM(dbp, page);
sumlen = pgsz;
break;
}
if (F_ISSET(dbp, DB_AM_SWAP))
P_32_SWAP(chksum);
if ((ret = __db_check_chksum_algo(dbenv, dbenv->crypto_handle,
(void *)chksum, page, sumlen,
is_hmac, algo)) != 0) {
logmsg(LOGMSG_ERROR, "pgno %u invalid checksum\n",
F_ISSET(dbp, DB_AM_SWAP) ? flibc_intflip(page->pgno)
: page->pgno);
continue;
}
}
if (is_hmac) {
DB_CIPHER *db_cipher = dbenv->crypto_handle;
void *iv = P_IV(dbp, page);
size_t skip = P_OVERHEAD(dbp);
uint8_t *ciphertext = (uint8_t *)page + skip;
if ((ret = db_cipher->decrypt(dbenv, db_cipher->data, iv,
ciphertext, sumlen - skip)) != 0) {
logmsg(LOGMSG_ERROR, "pgno %u decryption failed\n", page->pgno);
continue;
}
}
if (IS_PREFIX(page) && F_ISSET(dbp, DB_AM_SWAP))
prefix_tocpu(dbp, page);
db_indx_t n = NUM_ENT(page);
if (F_ISSET(dbp, DB_AM_SWAP))
n = flibc_shortflip(n);
if (n == 0)
continue;
/* We only care about the key so we take half entries
on the page. We don't check the flags of every entry
to get the count, so it's likely deleted entries are
counted here. However this is okay as we only need these
two counters to estimate the number of periodic stat4 samples.
And because entries may be deleted after we check them,
even if we did check every entry, the results wouldn't be
100% accurate anyway. */
recs_looked_at += (n >> 1);
if (rand() % 100 >= comp_pct)
continue;
NUM_ENT(page) = n;
nrecs += (n >> 1);
/* Check disk space, schema changes, analyze abort request etc. */
now = comdb2_time_epoch();
if (now - last >= 10) {
last = now;
rc = check_free_space(bdb_state->dir);
if (rc != BDBERR_NOERROR) {
*bdberr = rc;
rc = -1;
goto done;
}
}
int inprogress;
if ((inprogress = get_schema_change_in_progress(__func__, __LINE__)) ||
get_analyze_abort_requested() || db_is_exiting()) {
if (inprogress)
logmsg(LOGMSG_ERROR, "%s: Aborting Analyze because "
"schema_change_in_progress\n", __func__);
if (get_analyze_abort_requested())
logmsg(LOGMSG_ERROR, "%s: Aborting Analyze because "
"of send analyze abort\n", __func__);
if (db_is_exiting())
logmsg(LOGMSG_ERROR, "%s: Aborting Analyze because "
"db is exiting\n", __func__);
rc = -1;
goto done;
}
db_indx_t *inp = P_INP(dbp, page);
/* Remember the value before byteswap.
We need to reset inp[0] before
saving the page to the temptable. */
db_indx_t originp = inp[0];
if (F_ISSET(dbp, DB_AM_SWAP))
inp[0] = flibc_shortflip(inp[0]);
BKEYDATA *data = GET_BKEYDATA(dbp, page, 0);
assert((uint8_t *)data < max);
/* skip deleted */
if (B_DISSET(data))
continue;
if (B_TYPE(data) != B_KEYDATA)
continue;
/* Remember the values before byteswap.
We need to reset 1st entry before
saving the page to the temptable. */
BKEYDATA *origdta = data;
db_indx_t origdlen = data->len;
if (F_ISSET(dbp, DB_AM_SWAP))
data->len = flibc_shortflip(data->len);
db_indx_t len;
ASSIGN_ALIGN(db_indx_t, len, data->len);
assert(((uint8_t *)data + len) < max);
if (bk_decompress(dbp, page, &data, pfxbuf, sizeof(pfxbuf)) != 0) {
logmsg(LOGMSG_ERROR,
"\ndecompress failed page:%d indx:0 total:%d\n", page->pgno,
n);
continue;
}
ASSIGN_ALIGN(db_indx_t, len, data->len);
/* Reset the 1st index and entry. */
inp[0] = originp;
origdta->len = origdlen;
/* Save the entire page:
key is the 1st key on the page;
data is the page itself. */
rc = bdb_temp_table_put(bdb_state->parent, sampler->tmptbl, data->data,
len, page, pgsz, NULL, bdberr);
if (rc)
goto done;
}
if (rc != 0) {
logmsg(LOGMSG_ERROR, "Problem reading dta file: %d %s\n", errno,
strerror(errno));
rc = -1;
goto done;
}
logmsg(LOGMSG_INFO, "summarize added %llu records, traversed %llu\n", nrecs,
recs_looked_at);
done:
if (fd != -1)
close(fd);
if (page)
free(page);
if (rc || *bdberr != BDBERR_NOERROR) {
if (sampler && *samplerp == NULL)
sampler_close(sampler);
return rc;
}
sampler->db = dbp_;
*outrecs = nrecs;
*cmprecs = recs_looked_at;
if (*samplerp == NULL)
*samplerp = sampler;
return rc;
}