ctdb/server/ctdb_vacuum.c

   1 /*
   2    ctdb vacuuming events
   3
   4    Copyright (C) Ronnie Sahlberg  2009
   5
   6    This program is free software; you can redistribute it and/or modify
   7    it under the terms of the GNU General Public License as published by
   8    the Free Software Foundation; either version 3 of the License, or
   9    (at your option) any later version.
  10
  11    This program is distributed in the hope that it will be useful,
  12    but WITHOUT ANY WARRANTY; without even the implied warranty of
  13    MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
  14    GNU General Public License for more details.
  15
  16    You should have received a copy of the GNU General Public License
  17    along with this program; if not, see <http://www.gnu.org/licenses/>.
  18 */
  19
  20 #include "includes.h"
  21 #include "lib/tevent/tevent.h"
  22 #include "lib/tdb/include/tdb.h"
  23 #include "system/network.h"
  24 #include "system/filesys.h"
  25 #include "system/dir.h"
  26 #include "../include/ctdb_private.h"
  27 #include "db_wrap.h"
  28 #include "lib/util/dlinklist.h"
  29 #include "lib/tevent/tevent.h"
  30 #include "../include/ctdb_private.h"
  31 #include "../common/rb_tree.h"
  32
  33 #define TIMELIMIT() timeval_current_ofs(10, 0)
  34 #define TUNINGDBNAME "vactune.tdb"
  35
  36 enum vacuum_child_status { VACUUM_RUNNING, VACUUM_OK, VACUUM_ERROR, VACUUM_TIMEOUT};
  37
  38 struct ctdb_vacuum_child_context {
  39         struct ctdb_vacuum_child_context *next, *prev;
  40         struct ctdb_vacuum_handle *vacuum_handle;
  41         /* fd child writes status to */
  42         int fd[2];
  43         pid_t child_pid;
  44         enum vacuum_child_status status;
  45         struct timeval start_time;
  46 };
  47
  48 struct ctdb_vacuum_handle {
  49         struct ctdb_db_context *ctdb_db;
  50         struct ctdb_vacuum_child_context *child_ctx;
  51 };
  52
  53
  54 /*  a list of records to possibly delete */
  55 struct vacuum_data {
  56         uint32_t vacuum_limit;
  57         uint32_t repack_limit;
  58         struct ctdb_context *ctdb;
  59         struct ctdb_db_context *ctdb_db;
  60         struct tdb_context *dest_db;
  61         trbt_tree_t *delete_tree;
  62         uint32_t delete_count;
  63         struct ctdb_marshall_buffer **list;
  64         struct timeval start;
  65         bool traverse_error;
  66         bool vacuum;
  67         uint32_t total;
  68         uint32_t vacuumed;
  69         uint32_t copied;
  70 };
  71
  72 /* tuning information stored for every db */
  73 struct vacuum_tuning_data {
  74         uint32_t last_num_repack;
  75         uint32_t last_num_empty;
  76         uint32_t last_interval;
  77         uint32_t new_interval;
  78         struct timeval last_start;
  79         double   last_duration;
  80 };
  81
  82 /* this structure contains the information for one record to be deleted */
  83 struct delete_record_data {
  84         struct ctdb_context *ctdb;
  85         struct ctdb_db_context *ctdb_db;
  86         struct ctdb_ltdb_header hdr;
  87         TDB_DATA key;
  88 };
  89
  90 struct delete_records_list {
  91         struct ctdb_marshall_buffer *records;
  92 };
  93
  94
  95 static int add_record_to_delete_tree(struct vacuum_data *vdata, TDB_DATA key,
  96                                      struct ctdb_ltdb_header *hdr)
  97 {
  98         struct ctdb_context *ctdb = vdata->ctdb;
  99         struct ctdb_db_context *ctdb_db = vdata->ctdb_db;
 100         uint32_t hash;
 101         struct delete_record_data *dd;
 102
 103         hash = ctdb_hash(&key);
 104
 105         if (trbt_lookup32(vdata->delete_tree, hash)) {
 106                 DEBUG(DEBUG_DEBUG, (__location__ " Hash collission when vacuuming, skipping this record.\n"));
 107                 return 0;
 108         }
 109
 110         /* store key and header indexed by the key hash */
 111         dd = talloc_zero(vdata->delete_tree, struct delete_record_data);
 112         if (dd == NULL) {
 113                 DEBUG(DEBUG_ERR,(__location__ " Out of memory\n"));
 114                 return -1;
 115         }
 116         dd->ctdb      = ctdb;
 117         dd->ctdb_db   = ctdb_db;
 118         dd->key.dsize = key.dsize;
 119         dd->key.dptr  = talloc_memdup(dd, key.dptr, key.dsize);
 120         if (dd->key.dptr == NULL) {
 121                 DEBUG(DEBUG_ERR,(__location__ " Out of memory\n"));
 122                 return -1;
 123         }
 124
 125         dd->hdr = *hdr;
 126
 127         trbt_insert32(vdata->delete_tree, hash, dd);
 128
 129         vdata->delete_count++;
 130
 131         return 0;
 132 }
 133
 134 /**
 135  * Add a record to the list of records to be sent
 136  * to their lmaster with VACUUM_FETCH.
 137  */
 138 static int add_record_to_vacuum_fetch_list(struct vacuum_data *vdata,
 139                                            TDB_DATA key)
 140 {
 141         struct ctdb_context *ctdb = vdata->ctdb;
 142         struct ctdb_rec_data *rec;
 143         uint32_t lmaster;
 144         size_t old_size;
 145
 146         lmaster = ctdb_lmaster(ctdb, &key);
 147
 148         rec = ctdb_marshall_record(vdata->list[lmaster], ctdb->pnn, key, NULL, tdb_null);
 149         if (rec == NULL) {
 150                 DEBUG(DEBUG_ERR,(__location__ " Out of memory\n"));
 151                 vdata->traverse_error = true;
 152                 return -1;
 153         }
 154
 155         old_size = talloc_get_size(vdata->list[lmaster]);
 156         vdata->list[lmaster] = talloc_realloc_size(NULL, vdata->list[lmaster],
 157                                                    old_size + rec->length);
 158         if (vdata->list[lmaster] == NULL) {
 159                 DEBUG(DEBUG_ERR,(__location__ " Failed to expand\n"));
 160                 vdata->traverse_error = true;
 161                 return -1;
 162         }
 163
 164         vdata->list[lmaster]->count++;
 165         memcpy(old_size+(uint8_t *)vdata->list[lmaster], rec, rec->length);
 166         talloc_free(rec);
 167
 168         vdata->total++;
 169
 170         return 0;
 171 }
 172
 173
 174 static void ctdb_vacuum_event(struct event_context *ev, struct timed_event *te,
 175                                                           struct timeval t, void *private_data);
 176
 177
 178 /*
 179  * traverse function for gathering the records that can be deleted
 180  */
 181 static int vacuum_traverse(struct tdb_context *tdb, TDB_DATA key, TDB_DATA data, void *private)
 182 {
 183         struct vacuum_data *vdata = talloc_get_type(private, struct vacuum_data);
 184         struct ctdb_context *ctdb = vdata->ctdb;
 185         uint32_t lmaster;
 186         struct ctdb_ltdb_header *hdr;
 187         int res = 0;
 188
 189         lmaster = ctdb_lmaster(ctdb, &key);
 190         if (lmaster >= ctdb->num_nodes) {
 191                 DEBUG(DEBUG_CRIT, (__location__
 192                                    " lmaster[%u] >= ctdb->num_nodes[%u] for key"
 193                                    " with hash[%u]!\n",
 194                                    (unsigned)lmaster,
 195                                    (unsigned)ctdb->num_nodes,
 196                                    (unsigned)ctdb_hash(&key)));
 197                 return -1;
 198         }
 199
 200         if (data.dsize != sizeof(struct ctdb_ltdb_header)) {
 201                 /* its not a deleted record */
 202                 return 0;
 203         }
 204
 205         hdr = (struct ctdb_ltdb_header *)data.dptr;
 206
 207         if (hdr->dmaster != ctdb->pnn) {
 208                 return 0;
 209         }
 210
 211         if (lmaster == ctdb->pnn) {
 212                 /*
 213                  * We are both lmaster and dmaster, and the record * is empty.
 214                  * So we should be able to delete it.
 215                  */
 216                 res = add_record_to_delete_tree(vdata, key, hdr);
 217         } else {
 218                 /*
 219                  * We are not lmaster.
 220                  * Add the record to the blob ready to send to the nodes.
 221                  */
 222                 res = add_record_to_vacuum_fetch_list(vdata, key);
 223         }
 224
 225         return res;
 226 }
 227
 228 /*
 229  * traverse the tree of records to delete and marshall them into
 230  * a blob
 231  */
 232 static void delete_traverse(void *param, void *data)
 233 {
 234         struct delete_record_data *dd = talloc_get_type(data, struct delete_record_data);
 235         struct delete_records_list *recs = talloc_get_type(param, struct delete_records_list);
 236         struct ctdb_rec_data *rec;
 237         size_t old_size;
 238
 239         rec = ctdb_marshall_record(dd, recs->records->db_id, dd->key, &dd->hdr, tdb_null);
 240         if (rec == NULL) {
 241                 DEBUG(DEBUG_ERR, (__location__ " failed to marshall record\n"));
 242                 return;
 243         }
 244
 245         old_size = talloc_get_size(recs->records);
 246         recs->records = talloc_realloc_size(NULL, recs->records, old_size + rec->length);
 247         if (recs->records == NULL) {
 248                 DEBUG(DEBUG_ERR,(__location__ " Failed to expand\n"));
 249                 return;
 250         }
 251         recs->records->count++;
 252         memcpy(old_size+(uint8_t *)(recs->records), rec, rec->length);
 253 }
 254
 255 /*
 256  * read-only traverse the database in order to find
 257  * records that can be deleted and try to delete these
 258  * records on the other nodes
 259  * this executes in the child context
 260  */
 261 static int ctdb_vacuum_db(struct ctdb_db_context *ctdb_db, struct vacuum_data *vdata)
 262 {
 263         struct ctdb_context *ctdb = ctdb_db->ctdb;
 264         const char *name = ctdb_db->db_name;
 265         int ret, i, pnn;
 266
 267         ret = ctdb_ctrl_getvnnmap(ctdb, TIMELIMIT(), CTDB_CURRENT_NODE, ctdb, &ctdb->vnn_map);
 268         if (ret != 0) {
 269                 DEBUG(DEBUG_ERR, ("Unable to get vnnmap from local node\n"));
 270                 return ret;
 271         }
 272
 273         pnn = ctdb_ctrl_getpnn(ctdb, TIMELIMIT(), CTDB_CURRENT_NODE);
 274         if (pnn == -1) {
 275                 DEBUG(DEBUG_ERR, ("Unable to get pnn from local node\n"));
 276                 return -1;
 277         }
 278
 279         ctdb->pnn = pnn;
 280         /* the list needs to be of length num_nodes */
 281         vdata->list = talloc_array(vdata, struct ctdb_marshall_buffer *, ctdb->num_nodes);
 282         if (vdata->list == NULL) {
 283                 DEBUG(DEBUG_ERR,(__location__ " Out of memory\n"));
 284                 return -1;
 285         }
 286         for (i = 0; i < ctdb->num_nodes; i++) {
 287                 vdata->list[i] = (struct ctdb_marshall_buffer *)
 288                         talloc_zero_size(vdata->list,
 289                                                          offsetof(struct ctdb_marshall_buffer, data));
 290                 if (vdata->list[i] == NULL) {
 291                         DEBUG(DEBUG_ERR,(__location__ " Out of memory\n"));
 292                         return -1;
 293                 }
 294                 vdata->list[i]->db_id = ctdb_db->db_id;
 295         }
 296
 297         /* read-only traverse, looking for records that might be able to be vacuumed */
 298         if (tdb_traverse_read(ctdb_db->ltdb->tdb, vacuum_traverse, vdata) == -1 ||
 299             vdata->traverse_error) {
 300                 DEBUG(DEBUG_ERR,(__location__ " Traverse error in vacuuming '%s'\n", name));
 301                 return -1;
 302         }
 303
 304         /*
 305          * For records where we are not the lmaster,
 306          * tell the lmaster to fetch the record.
 307          */
 308         for (i = 0; i < ctdb->num_nodes; i++) {
 309                 TDB_DATA data;
 310
 311                 if (ctdb->nodes[i]->pnn == ctdb->pnn) {
 312                         continue;
 313                 }
 314
 315                 if (vdata->list[i]->count == 0) {
 316                         continue;
 317                 }
 318
 319                 DEBUG(DEBUG_INFO, ("Found %u records for lmaster %u in '%s'\n",
 320                                    vdata->list[i]->count, ctdb->nodes[i]->pnn,
 321                                    name));
 322
 323                 data.dsize = talloc_get_size(vdata->list[i]);
 324                 data.dptr  = (void *)vdata->list[i];
 325                 if (ctdb_client_send_message(ctdb, ctdb->nodes[i]->pnn, CTDB_SRVID_VACUUM_FETCH, data) != 0) {
 326                         DEBUG(DEBUG_ERR, (__location__ " Failed to send vacuum "
 327                                           "fetch message to %u\n",
 328                                           ctdb->nodes[i]->pnn));
 329                         return -1;
 330                 }
 331         }
 332
 333         /* Process all records we can delete (if any) */
 334         if (vdata->delete_count > 0) {
 335                 struct delete_records_list *recs;
 336                 TDB_DATA indata, outdata;
 337                 int32_t res;
 338                 struct ctdb_node_map *nodemap;
 339                 uint32_t *active_nodes;
 340                 int num_active_nodes;
 341
 342                 recs = talloc_zero(vdata, struct delete_records_list);
 343                 if (recs == NULL) {
 344                         DEBUG(DEBUG_ERR,(__location__ " Out of memory\n"));
 345                         return -1;
 346                 }
 347                 recs->records = (struct ctdb_marshall_buffer *)
 348                         talloc_zero_size(vdata,
 349                                     offsetof(struct ctdb_marshall_buffer, data));
 350                 if (recs->records == NULL) {
 351                         DEBUG(DEBUG_ERR,(__location__ " Out of memory\n"));
 352                         return -1;
 353                 }
 354                 recs->records->db_id = ctdb_db->db_id;
 355
 356                 /*
 357                  * traverse the tree of all records we want to delete and
 358                  * create a blob we can send to the other nodes.
 359                  */
 360                 trbt_traversearray32(vdata->delete_tree, 1, delete_traverse, recs);
 361
 362                 indata.dsize = talloc_get_size(recs->records);
 363                 indata.dptr  = (void *)recs->records;
 364
 365                 /*
 366                  * now tell all the active nodes to delete all these records
 367                  * (if possible)
 368                  */
 369
 370                 ret = ctdb_ctrl_getnodemap(ctdb, TIMELIMIT(),
 371                                            CTDB_CURRENT_NODE,
 372                                            recs, /* talloc context */
 373                                            &nodemap);
 374                 if (ret != 0) {
 375                         DEBUG(DEBUG_ERR,(__location__ " unable to get node map\n"));
 376                         return -1;
 377                 }
 378
 379                 active_nodes = list_of_active_nodes(ctdb, nodemap,
 380                                                     nodemap, /* talloc context */
 381                                                     false /* include self */);
 382                 /* yuck! ;-) */
 383                 num_active_nodes = talloc_get_size(active_nodes)/sizeof(*active_nodes);
 384
 385                 for (i = 0; i < num_active_nodes; i++) {
 386                         struct ctdb_marshall_buffer *records;
 387                         struct ctdb_rec_data *rec;
 388
 389                         ret = ctdb_control(ctdb, active_nodes[i], 0,
 390                                         CTDB_CONTROL_TRY_DELETE_RECORDS, 0,
 391                                         indata, recs, &outdata, &res,
 392                                         NULL, NULL);
 393                         if (ret != 0 || res != 0) {
 394                                 DEBUG(DEBUG_ERR, ("Failed to delete records on "
 395                                                   "node %u: ret[%d] res[%d]\n",
 396                                                   active_nodes[i], ret, res));
 397                                 return -1;
 398                         }
 399
 400                         /*
 401                          * outdata countains the list of records coming back
 402                          * from the node which the node could not delete
 403                          */
 404                         records = (struct ctdb_marshall_buffer *)outdata.dptr;
 405                         rec = (struct ctdb_rec_data *)&records->data[0];
 406                         while (records->count-- > 1) {
 407                                 TDB_DATA reckey, recdata;
 408                                 struct ctdb_ltdb_header *rechdr;
 409
 410                                 reckey.dptr = &rec->data[0];
 411                                 reckey.dsize = rec->keylen;
 412                                 recdata.dptr = &rec->data[reckey.dsize];
 413                                 recdata.dsize = rec->datalen;
 414
 415                                 if (recdata.dsize < sizeof(struct ctdb_ltdb_header)) {
 416                                         DEBUG(DEBUG_CRIT,(__location__ " bad ltdb record\n"));
 417                                         return -1;
 418                                 }
 419                                 rechdr = (struct ctdb_ltdb_header *)recdata.dptr;
 420                                 recdata.dptr += sizeof(*rechdr);
 421                                 recdata.dsize -= sizeof(*rechdr);
 422
 423                                 /*
 424                                  * that other node couldnt delete the record
 425                                  * so we should delete it and thereby remove it from the tree
 426                                  */
 427                                 talloc_free(trbt_lookup32(vdata->delete_tree, ctdb_hash(&reckey)));
 428
 429                                 rec = (struct ctdb_rec_data *)(rec->length + (uint8_t *)rec);
 430                         }
 431                 }
 432
 433                 /* free nodemap and active_nodes */
 434                 talloc_free(nodemap);
 435
 436                 /*
 437                  * The only records remaining in the tree would be those
 438                  * records where all other nodes could successfully
 439                  * delete them, so we can safely delete them on the
 440                  * lmaster as well. Deletion implictely happens while
 441                  * we repack the database. The repack algorithm revisits
 442                  * the tree in order to find the records that don't need
 443                  * to be copied / repacked.
 444                  */
 445         }
 446
 447         /* this ensures we run our event queue */
 448         ctdb_ctrl_getpnn(ctdb, TIMELIMIT(), CTDB_CURRENT_NODE);
 449
 450         return 0;
 451 }
 452
 453
 454 /*
 455  * traverse function for repacking
 456  */
 457 static int repack_traverse(struct tdb_context *tdb, TDB_DATA key, TDB_DATA data, void *private)
 458 {
 459         struct vacuum_data *vdata = (struct vacuum_data *)private;
 460
 461         if (vdata->vacuum) {
 462                 uint32_t hash = ctdb_hash(&key);
 463                 struct delete_record_data *kd;
 464                 /*
 465                  * check if we can ignore this record because it's in the delete_tree
 466                  */
 467                 kd = (struct delete_record_data *)trbt_lookup32(vdata->delete_tree, hash);
 468                 /*
 469                  * there might be hash collisions so we have to compare the keys here to be sure
 470                  */
 471                 if (kd && kd->key.dsize == key.dsize && memcmp(kd->key.dptr, key.dptr, key.dsize) == 0) {
 472                         struct ctdb_ltdb_header *hdr = (struct ctdb_ltdb_header *)data.dptr;
 473                         /*
 474                          * we have to check if the record hasn't changed in the meantime in order to
 475                          * savely remove it from the database
 476                          */
 477                         if (data.dsize == sizeof(struct ctdb_ltdb_header) &&
 478                                 hdr->dmaster == kd->ctdb->pnn &&
 479                                 ctdb_lmaster(kd->ctdb, &(kd->key)) == kd->ctdb->pnn &&
 480                                 kd->hdr.rsn == hdr->rsn) {
 481                                 vdata->vacuumed++;
 482                                 return 0;
 483                         }
 484                 }
 485         }
 486         if (tdb_store(vdata->dest_db, key, data, TDB_INSERT) != 0) {
 487                 vdata->traverse_error = true;
 488                 return -1;
 489         }
 490         vdata->copied++;
 491         return 0;
 492 }
 493
 494 /*
 495  * repack a tdb
 496  */
 497 static int ctdb_repack_tdb(struct tdb_context *tdb, TALLOC_CTX *mem_ctx, struct vacuum_data *vdata)
 498 {
 499         struct tdb_context *tmp_db;
 500
 501         if (tdb_transaction_start(tdb) != 0) {
 502                 DEBUG(DEBUG_ERR,(__location__ " Failed to start transaction\n"));
 503                 return -1;
 504         }
 505
 506         tmp_db = tdb_open("tmpdb", tdb_hash_size(tdb),
 507                           TDB_INTERNAL|TDB_DISALLOW_NESTING,
 508                           O_RDWR|O_CREAT, 0);
 509         if (tmp_db == NULL) {
 510                 DEBUG(DEBUG_ERR,(__location__ " Failed to create tmp_db\n"));
 511                 tdb_transaction_cancel(tdb);
 512                 return -1;
 513         }
 514
 515         vdata->traverse_error = false;
 516         vdata->dest_db = tmp_db;
 517         vdata->vacuum = true;
 518         vdata->vacuumed = 0;
 519         vdata->copied = 0;
 520
 521         /*
 522          * repack and vacuum on-the-fly by not writing the records that are
 523          * no longer needed
 524          */
 525         if (tdb_traverse_read(tdb, repack_traverse, vdata) == -1) {
 526                 DEBUG(DEBUG_ERR,(__location__ " Failed to traverse copying out\n"));
 527                 tdb_transaction_cancel(tdb);
 528                 tdb_close(tmp_db);
 529                 return -1;
 530         }
 531
 532         DEBUG(DEBUG_INFO,(__location__ " %u records vacuumed\n", vdata->vacuumed));
 533
 534         if (vdata->traverse_error) {
 535                 DEBUG(DEBUG_ERR,(__location__ " Error during traversal\n"));
 536                 tdb_transaction_cancel(tdb);
 537                 tdb_close(tmp_db);
 538                 return -1;
 539         }
 540
 541         if (tdb_wipe_all(tdb) != 0) {
 542                 DEBUG(DEBUG_ERR,(__location__ " Failed to wipe database\n"));
 543                 tdb_transaction_cancel(tdb);
 544                 tdb_close(tmp_db);
 545                 return -1;
 546         }
 547
 548         vdata->traverse_error = false;
 549         vdata->dest_db = tdb;
 550         vdata->vacuum = false;
 551         vdata->copied = 0;
 552
 553         if (tdb_traverse_read(tmp_db, repack_traverse, vdata) == -1) {
 554                 DEBUG(DEBUG_ERR,(__location__ " Failed to traverse copying back\n"));
 555                 tdb_transaction_cancel(tdb);
 556                 tdb_close(tmp_db);
 557                 return -1;
 558         }
 559
 560         if (vdata->traverse_error) {
 561                 DEBUG(DEBUG_ERR,(__location__ " Error during second traversal\n"));
 562                 tdb_transaction_cancel(tdb);
 563                 tdb_close(tmp_db);
 564                 return -1;
 565         }
 566
 567         tdb_close(tmp_db);
 568
 569
 570         if (tdb_transaction_commit(tdb) != 0) {
 571                 DEBUG(DEBUG_ERR,(__location__ " Failed to commit\n"));
 572                 return -1;
 573         }
 574         DEBUG(DEBUG_INFO,(__location__ " %u records copied\n", vdata->copied));
 575
 576         return 0;
 577 }
 578
 579 static int update_tuning_db(struct ctdb_db_context *ctdb_db, struct vacuum_data *vdata, uint32_t freelist)
 580 {
 581         TALLOC_CTX *tmp_ctx = talloc_new(NULL);
 582         TDB_CONTEXT *tune_tdb;
 583         TDB_DATA key, value;
 584         struct vacuum_tuning_data tdata;
 585         struct vacuum_tuning_data *tptr;
 586         char *vac_dbname;
 587         int flags;
 588
 589         vac_dbname = talloc_asprintf(tmp_ctx, "%s/%s.%u",
 590                                      ctdb_db->ctdb->db_directory_state,
 591                                      TUNINGDBNAME, ctdb_db->ctdb->pnn);
 592         if (vac_dbname == NULL) {
 593                 DEBUG(DEBUG_CRIT,(__location__ " Out of memory error while allocating '%s'\n", vac_dbname));
 594                 talloc_free(tmp_ctx);
 595                 return -1;
 596         }
 597
 598         flags  = ctdb_db->ctdb->valgrinding ? TDB_NOMMAP : 0;
 599         flags |= TDB_DISALLOW_NESTING;
 600         tune_tdb = tdb_open(vac_dbname, 0,
 601                             flags,
 602                             O_RDWR|O_CREAT, 0600);
 603         if (tune_tdb == NULL) {
 604                 DEBUG(DEBUG_ERR,(__location__ " Failed to create/open %s\n", TUNINGDBNAME));
 605                 talloc_free(tmp_ctx);
 606                 return -1;
 607         }
 608
 609         if (tdb_transaction_start(tune_tdb) != 0) {
 610                 DEBUG(DEBUG_ERR,(__location__ " Failed to start transaction\n"));
 611                 tdb_close(tune_tdb);
 612                 return -1;
 613         }
 614         key.dptr = discard_const(ctdb_db->db_name);
 615         key.dsize = strlen(ctdb_db->db_name);
 616         value = tdb_fetch(tune_tdb, key);
 617
 618         if (value.dptr != NULL && value.dsize == sizeof(struct vacuum_tuning_data)) {
 619                 tptr = (struct vacuum_tuning_data *)value.dptr;
 620                 tdata = *tptr;
 621
 622                 /*
 623                  * re-calc new vacuum interval:
 624                  * in case no limit was reached we continously increase the interval
 625                  * until vacuum_max_interval is reached
 626                  * in case a limit was reached we divide the current interval by 2
 627                  * unless vacuum_min_interval is reached
 628                  */
 629                 if (freelist < vdata->repack_limit &&
 630                     vdata->delete_count < vdata->vacuum_limit) {
 631                         if (tdata.last_interval < ctdb_db->ctdb->tunable.vacuum_max_interval) {
 632                                 tdata.new_interval = tdata.last_interval * 110 / 100;
 633                                 DEBUG(DEBUG_INFO,("Increasing vacuum interval %u -> %u for %s\n",
 634                                         tdata.last_interval, tdata.new_interval, ctdb_db->db_name));
 635                         }
 636                 } else {
 637                         tdata.new_interval = tdata.last_interval / 2;
 638                         if (tdata.new_interval < ctdb_db->ctdb->tunable.vacuum_min_interval ||
 639                                 tdata.new_interval > ctdb_db->ctdb->tunable.vacuum_max_interval) {
 640                                 tdata.new_interval = ctdb_db->ctdb->tunable.vacuum_min_interval;
 641                         }
 642                         DEBUG(DEBUG_INFO,("Decreasing vacuum interval %u -> %u for %s\n",
 643                                          tdata.last_interval, tdata.new_interval, ctdb_db->db_name));
 644                 }
 645                 tdata.last_interval = tdata.new_interval;
 646         } else {
 647                 DEBUG(DEBUG_DEBUG,(__location__ " Cannot find tunedb record for %s. Using default interval\n", ctdb_db->db_name));
 648                 tdata.last_num_repack = freelist;
 649                 tdata.last_num_empty = vdata->delete_count;
 650                 tdata.last_interval = ctdb_db->ctdb->tunable.vacuum_default_interval;
 651         }
 652
 653         if (value.dptr != NULL) {
 654                 free(value.dptr);
 655         }
 656
 657         tdata.last_start = vdata->start;
 658         tdata.last_duration = timeval_elapsed(&vdata->start);
 659
 660         value.dptr = (unsigned char *)&tdata;
 661         value.dsize = sizeof(tdata);
 662
 663         if (tdb_store(tune_tdb, key, value, 0) != 0) {
 664                 DEBUG(DEBUG_ERR,(__location__ " Unable to store tundb record for %s\n", ctdb_db->db_name));
 665                 tdb_transaction_cancel(tune_tdb);
 666                 tdb_close(tune_tdb);
 667                 talloc_free(tmp_ctx);
 668                 return -1;
 669         }
 670         tdb_transaction_commit(tune_tdb);
 671         tdb_close(tune_tdb);
 672         talloc_free(tmp_ctx);
 673
 674         return 0;
 675 }
 676
 677 /*
 678  * repack and vaccum a db
 679  * called from the child context
 680  */
 681 static int ctdb_vacuum_and_repack_db(struct ctdb_db_context *ctdb_db,
 682                                      TALLOC_CTX *mem_ctx)
 683 {
 684         uint32_t repack_limit = ctdb_db->ctdb->tunable.repack_limit;
 685         uint32_t vacuum_limit = ctdb_db->ctdb->tunable.vacuum_limit;
 686         const char *name = ctdb_db->db_name;
 687         int size;
 688         struct vacuum_data *vdata;
 689
 690         size = tdb_freelist_size(ctdb_db->ltdb->tdb);
 691         if (size == -1) {
 692                 DEBUG(DEBUG_ERR,(__location__ " Failed to get freelist size for '%s'\n", name));
 693                 return -1;
 694         }
 695
 696         vdata = talloc_zero(mem_ctx, struct vacuum_data);
 697         if (vdata == NULL) {
 698                 DEBUG(DEBUG_ERR,(__location__ " Out of memory\n"));
 699                 return -1;
 700         }
 701
 702         vdata->ctdb = ctdb_db->ctdb;
 703         vdata->vacuum_limit = vacuum_limit;
 704         vdata->repack_limit = repack_limit;
 705         vdata->delete_tree = trbt_create(vdata, 0);
 706         vdata->ctdb_db = ctdb_db;
 707         if (vdata->delete_tree == NULL) {
 708                 DEBUG(DEBUG_ERR,(__location__ " Out of memory\n"));
 709                 talloc_free(vdata);
 710                 return -1;
 711         }
 712
 713         vdata->start = timeval_current();
 714
 715         /*
 716          * gather all records that can be deleted in vdata
 717          */
 718         if (ctdb_vacuum_db(ctdb_db, vdata) != 0) {
 719                 DEBUG(DEBUG_ERR,(__location__ " Failed to vacuum '%s'\n", name));
 720         }
 721
 722         /*
 723          * decide if a repack is necessary
 724          */
 725         if (size < repack_limit && vdata->delete_count < vacuum_limit) {
 726                 update_tuning_db(ctdb_db, vdata, size);
 727                 talloc_free(vdata);
 728                 return 0;
 729         }
 730
 731         DEBUG(DEBUG_INFO,("Repacking %s with %u freelist entries and %u records to delete\n",
 732                         name, size, vdata->delete_count));
 733
 734         /*
 735          * repack and implicitely get rid of the records we can delete
 736          */
 737         if (ctdb_repack_tdb(ctdb_db->ltdb->tdb, mem_ctx, vdata) != 0) {
 738                 DEBUG(DEBUG_ERR,(__location__ " Failed to repack '%s'\n", name));
 739                 update_tuning_db(ctdb_db, vdata, size);
 740                 talloc_free(vdata);
 741                 return -1;
 742         }
 743         update_tuning_db(ctdb_db, vdata, size);
 744         talloc_free(vdata);
 745
 746         return 0;
 747 }
 748
 749 static int get_vacuum_interval(struct ctdb_db_context *ctdb_db)
 750 {
 751         TALLOC_CTX *tmp_ctx = talloc_new(NULL);
 752         TDB_CONTEXT *tdb;
 753         TDB_DATA key, value;
 754         char *vac_dbname;
 755         uint interval = ctdb_db->ctdb->tunable.vacuum_default_interval;
 756         struct ctdb_context *ctdb = ctdb_db->ctdb;
 757         int flags;
 758
 759         vac_dbname = talloc_asprintf(tmp_ctx, "%s/%s.%u", ctdb->db_directory, TUNINGDBNAME, ctdb->pnn);
 760         if (vac_dbname == NULL) {
 761                 DEBUG(DEBUG_CRIT,(__location__ " Out of memory error while allocating '%s'\n", vac_dbname));
 762                 talloc_free(tmp_ctx);
 763                 return interval;
 764         }
 765
 766         flags  = ctdb_db->ctdb->valgrinding ? TDB_NOMMAP : 0;
 767         flags |= TDB_DISALLOW_NESTING;
 768         tdb = tdb_open(vac_dbname, 0,
 769                        flags,
 770                        O_RDWR|O_CREAT, 0600);
 771         if (!tdb) {
 772                 DEBUG(DEBUG_ERR,("Unable to open/create database %s using default interval. Errno : %s (%d)\n", vac_dbname, strerror(errno), errno));
 773                 talloc_free(tmp_ctx);
 774                 return interval;
 775         }
 776
 777         key.dptr = discard_const(ctdb_db->db_name);
 778         key.dsize = strlen(ctdb_db->db_name);
 779
 780         value = tdb_fetch(tdb, key);
 781
 782         if (value.dptr != NULL) {
 783                 if (value.dsize == sizeof(struct vacuum_tuning_data)) {
 784                         struct vacuum_tuning_data *tptr = (struct vacuum_tuning_data *)value.dptr;
 785
 786                         interval = tptr->new_interval;
 787
 788                         if (interval < ctdb->tunable.vacuum_min_interval) {
 789                                 interval = ctdb->tunable.vacuum_min_interval;
 790                         }
 791                         if (interval > ctdb->tunable.vacuum_max_interval) {
 792                                 interval = ctdb->tunable.vacuum_max_interval;
 793                         }
 794                 }
 795                 free(value.dptr);
 796         }
 797         tdb_close(tdb);
 798
 799         talloc_free(tmp_ctx);
 800
 801         return interval;
 802 }
 803
 804 static int vacuum_child_destructor(struct ctdb_vacuum_child_context *child_ctx)
 805 {
 806         double l = timeval_elapsed(&child_ctx->start_time);
 807         struct ctdb_db_context *ctdb_db = child_ctx->vacuum_handle->ctdb_db;
 808         struct ctdb_context *ctdb = ctdb_db->ctdb;
 809
 810         DEBUG(DEBUG_INFO,("Vacuuming took %.3f seconds for database %s\n", l, ctdb_db->db_name));
 811
 812         if (child_ctx->child_pid != -1) {
 813                 kill(child_ctx->child_pid, SIGKILL);
 814         }
 815
 816         DLIST_REMOVE(ctdb->vacuumers, child_ctx);
 817
 818         event_add_timed(ctdb->ev, child_ctx->vacuum_handle,
 819                         timeval_current_ofs(get_vacuum_interval(ctdb_db), 0),
 820                         ctdb_vacuum_event, child_ctx->vacuum_handle);
 821
 822         return 0;
 823 }
 824
 825 /*
 826  * this event is generated when a vacuum child process times out
 827  */
 828 static void vacuum_child_timeout(struct event_context *ev, struct timed_event *te,
 829                                          struct timeval t, void *private_data)
 830 {
 831         struct ctdb_vacuum_child_context *child_ctx = talloc_get_type(private_data, struct ctdb_vacuum_child_context);
 832
 833         DEBUG(DEBUG_ERR,("Vacuuming child process timed out for db %s\n", child_ctx->vacuum_handle->ctdb_db->db_name));
 834
 835         child_ctx->status = VACUUM_TIMEOUT;
 836
 837         talloc_free(child_ctx);
 838 }
 839
 840
 841 /*
 842  * this event is generated when a vacuum child process has completed
 843  */
 844 static void vacuum_child_handler(struct event_context *ev, struct fd_event *fde,
 845                              uint16_t flags, void *private_data)
 846 {
 847         struct ctdb_vacuum_child_context *child_ctx = talloc_get_type(private_data, struct ctdb_vacuum_child_context);
 848         char c = 0;
 849         int ret;
 850
 851         DEBUG(DEBUG_INFO,("Vacuuming child process %d finished for db %s\n", child_ctx->child_pid, child_ctx->vacuum_handle->ctdb_db->db_name));
 852         child_ctx->child_pid = -1;
 853
 854         ret = read(child_ctx->fd[0], &c, 1);
 855         if (ret != 1 || c != 0) {
 856                 child_ctx->status = VACUUM_ERROR;
 857                 DEBUG(DEBUG_ERR, ("A vacuum child process failed with an error for database %s. ret=%d c=%d\n", child_ctx->vacuum_handle->ctdb_db->db_name, ret, c));
 858         } else {
 859                 child_ctx->status = VACUUM_OK;
 860         }
 861
 862         talloc_free(child_ctx);
 863 }
 864
 865 /*
 866  * this event is called every time we need to start a new vacuum process
 867  */
 868 static void
 869 ctdb_vacuum_event(struct event_context *ev, struct timed_event *te,
 870                                struct timeval t, void *private_data)
 871 {
 872         struct ctdb_vacuum_handle *vacuum_handle = talloc_get_type(private_data, struct ctdb_vacuum_handle);
 873         struct ctdb_db_context *ctdb_db = vacuum_handle->ctdb_db;
 874         struct ctdb_context *ctdb = ctdb_db->ctdb;
 875         struct ctdb_vacuum_child_context *child_ctx;
 876         struct tevent_fd *fde;
 877         int ret;
 878
 879         /* we dont vacuum if we are in recovery mode, or db frozen */
 880         if (ctdb->recovery_mode == CTDB_RECOVERY_ACTIVE ||
 881             ctdb->freeze_mode[ctdb_db->priority] != CTDB_FREEZE_NONE) {
 882                 DEBUG(DEBUG_INFO, ("Not vacuuming %s (%s)\n", ctdb_db->db_name,
 883                                    ctdb->recovery_mode == CTDB_RECOVERY_ACTIVE ? "in recovery"
 884                                    : ctdb->freeze_mode[ctdb_db->priority] == CTDB_FREEZE_PENDING
 885                                    ? "freeze pending"
 886                                    : "frozen"));
 887                 event_add_timed(ctdb->ev, vacuum_handle, timeval_current_ofs(ctdb->tunable.vacuum_default_interval, 0), ctdb_vacuum_event, vacuum_handle);
 888                 return;
 889         }
 890
 891         child_ctx = talloc(vacuum_handle, struct ctdb_vacuum_child_context);
 892         if (child_ctx == NULL) {
 893                 DEBUG(DEBUG_CRIT, (__location__ " Failed to allocate child context for vacuuming of %s\n", ctdb_db->db_name));
 894                 ctdb_fatal(ctdb, "Out of memory when crating vacuum child context. Shutting down\n");
 895         }
 896
 897
 898         ret = pipe(child_ctx->fd);
 899         if (ret != 0) {
 900                 talloc_free(child_ctx);
 901                 DEBUG(DEBUG_ERR, ("Failed to create pipe for vacuum child process.\n"));
 902                 event_add_timed(ctdb->ev, vacuum_handle, timeval_current_ofs(ctdb->tunable.vacuum_default_interval, 0), ctdb_vacuum_event, vacuum_handle);
 903                 return;
 904         }
 905
 906         child_ctx->child_pid = ctdb_fork(ctdb);
 907         if (child_ctx->child_pid == (pid_t)-1) {
 908                 close(child_ctx->fd[0]);
 909                 close(child_ctx->fd[1]);
 910                 talloc_free(child_ctx);
 911                 DEBUG(DEBUG_ERR, ("Failed to fork vacuum child process.\n"));
 912                 event_add_timed(ctdb->ev, vacuum_handle, timeval_current_ofs(ctdb->tunable.vacuum_default_interval, 0), ctdb_vacuum_event, vacuum_handle);
 913                 return;
 914         }
 915
 916
 917         if (child_ctx->child_pid == 0) {
 918                 char cc = 0;
 919                 close(child_ctx->fd[0]);
 920
 921                 DEBUG(DEBUG_INFO,("Vacuuming child process %d for db %s started\n", getpid(), ctdb_db->db_name));
 922
 923                 if (switch_from_server_to_client(ctdb, "vacuum-%s", ctdb_db->db_name) != 0) {
 924                         DEBUG(DEBUG_CRIT, (__location__ "ERROR: failed to switch vacuum daemon into client mode. Shutting down.\n"));
 925                         _exit(1);
 926                 }
 927
 928                 /*
 929                  * repack the db
 930                  */
 931                 cc = ctdb_vacuum_and_repack_db(ctdb_db, child_ctx);
 932
 933                 write(child_ctx->fd[1], &cc, 1);
 934                 _exit(0);
 935         }
 936
 937         set_close_on_exec(child_ctx->fd[0]);
 938         close(child_ctx->fd[1]);
 939
 940         child_ctx->status = VACUUM_RUNNING;
 941         child_ctx->start_time = timeval_current();
 942
 943         DLIST_ADD(ctdb->vacuumers, child_ctx);
 944         talloc_set_destructor(child_ctx, vacuum_child_destructor);
 945
 946         /*
 947          * Clear the fastpath vacuuming list in the parent.
 948          */
 949         talloc_free(ctdb_db->delete_queue);
 950         ctdb_db->delete_queue = trbt_create(ctdb_db, 0);
 951         if (ctdb_db->delete_queue == NULL) {
 952                 /* fatal here? ... */
 953                 ctdb_fatal(ctdb, "Out of memory when re-creating vacuum tree "
 954                                  "in parent context. Shutting down\n");
 955         }
 956
 957         event_add_timed(ctdb->ev, child_ctx,
 958                 timeval_current_ofs(ctdb->tunable.vacuum_max_run_time, 0),
 959                 vacuum_child_timeout, child_ctx);
 960
 961         DEBUG(DEBUG_DEBUG, (__location__ " Created PIPE FD:%d to child vacuum process\n", child_ctx->fd[0]));
 962
 963         fde = event_add_fd(ctdb->ev, child_ctx, child_ctx->fd[0],
 964                            EVENT_FD_READ, vacuum_child_handler, child_ctx);
 965         tevent_fd_set_auto_close(fde);
 966
 967         vacuum_handle->child_ctx = child_ctx;
 968         child_ctx->vacuum_handle = vacuum_handle;
 969 }
 970
 971 void ctdb_stop_vacuuming(struct ctdb_context *ctdb)
 972 {
 973         /* Simply free them all. */
 974         while (ctdb->vacuumers) {
 975                 DEBUG(DEBUG_INFO, ("Aborting vacuuming for %s (%i)\n",
 976                            ctdb->vacuumers->vacuum_handle->ctdb_db->db_name,
 977                            (int)ctdb->vacuumers->child_pid));
 978                 /* vacuum_child_destructor kills it, removes from list */
 979                 talloc_free(ctdb->vacuumers);
 980         }
 981 }
 982
 983 /* this function initializes the vacuuming context for a database
 984  * starts the vacuuming events
 985  */
 986 int ctdb_vacuum_init(struct ctdb_db_context *ctdb_db)
 987 {
 988         if (ctdb_db->persistent != 0) {
 989                 DEBUG(DEBUG_ERR,("Vacuuming is disabled for persistent database %s\n", ctdb_db->db_name));
 990                 return 0;
 991         }
 992
 993         ctdb_db->vacuum_handle = talloc(ctdb_db, struct ctdb_vacuum_handle);
 994         CTDB_NO_MEMORY(ctdb_db->ctdb, ctdb_db->vacuum_handle);
 995
 996         ctdb_db->vacuum_handle->ctdb_db = ctdb_db;
 997
 998         event_add_timed(ctdb_db->ctdb->ev, ctdb_db->vacuum_handle,
 999                         timeval_current_ofs(get_vacuum_interval(ctdb_db), 0),
1000                         ctdb_vacuum_event, ctdb_db->vacuum_handle);
1001
1002         return 0;
1003 }