sqlite_aclk.c 26 KB

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101102103104105106107108109110111112113114115116117118119120121122123124125126127128129130131132133134135136137138139140141142143144145146147148149150151152153154155156157158159160161162163164165166167168169170171172173174175176177178179180181182183184185186187188189190191192193194195196197198199200201202203204205206207208209210211212213214215216217218219220221222223224225226227228229230231232233234235236237238239240241242243244245246247248249250251252253254255256257258259260261262263264265266267268269270271272273274275276277278279280281282283284285286287288289290291292293294295296297298299300301302303304305306307308309310311312313314315316317318319320321322323324325326327328329330331332333334335336337338339340341342343344345346347348349350351352353354355356357358359360361362363364365366367368369370371372373374375376377378379380381382383384385386387388389390391392393394395396397398399400401402403404405406407408409410411412413414415416417418419420421422423424425426427428429430431432433434435436437438439440441442443444445446447448449450451452453454455456457458459460461462463464465466467468469470471472473474475476477478479480481482483484485486487488489490491492493494495496497498499500501502503504505506507508509510511512513514515516517518519520521522523524525526527528529530531532533534535536537538539540541542543544545546547548549550551552553554555556557558559560561562563564565566567568569570571572573574575576577578579580581582583584585586587588589590591592593594595596597598599600601602603604605606607608609610611612613614615616617618619620621622623624625626627628629630631632633634635636637638639640641642643644645646647648649650651652653654655656657658659660661662663664665666667668669670671672673674675676677678679680681682683684685686687688689690691692693694695696697698699700701702703704705706707708709710711712713714715716717718719720721722723724725726727728729730731732733734735736737738739740741742743744745746747748749750751752753754755756757758759760761762763764765766767768769770771772773774775776777778779780781782783784785786
  1. // SPDX-License-Identifier: GPL-3.0-or-later
  2. #include "sqlite_functions.h"
  3. #include "sqlite_aclk.h"
  4. #include "sqlite_aclk_chart.h"
  5. #include "sqlite_aclk_node.h"
  6. #ifdef ENABLE_NEW_CLOUD_PROTOCOL
  7. #include "../../aclk/aclk.h"
  8. #endif
  9. const char *aclk_sync_config[] = {
  10. "CREATE TABLE IF NOT EXISTS dimension_delete (dimension_id blob, dimension_name text, chart_type_id text, "
  11. "dim_id blob, chart_id blob, host_id blob, date_created);",
  12. "CREATE INDEX IF NOT EXISTS ind_h1 ON dimension_delete (host_id);",
  13. "CREATE TRIGGER IF NOT EXISTS tr_dim_del AFTER DELETE ON dimension BEGIN INSERT INTO dimension_delete "
  14. "(dimension_id, dimension_name, chart_type_id, dim_id, chart_id, host_id, date_created)"
  15. " select old.id, old.name, c.type||\".\"||c.id, old.dim_id, old.chart_id, c.host_id, strftime('%s') FROM"
  16. " chart c WHERE c.chart_id = old.chart_id; END;",
  17. "DELETE FROM dimension_delete WHERE host_id NOT IN"
  18. " (SELECT host_id FROM host) OR strftime('%s') - date_created > 604800;",
  19. NULL,
  20. };
  21. uv_mutex_t aclk_async_lock;
  22. struct aclk_database_worker_config *aclk_thread_head = NULL;
  23. static inline int claimed()
  24. {
  25. int rc;
  26. rrdhost_aclk_state_lock(localhost);
  27. rc = (localhost->aclk_state.claimed_id != NULL);
  28. rrdhost_aclk_state_unlock(localhost);
  29. return rc;
  30. }
  31. void aclk_add_worker_thread(struct aclk_database_worker_config *wc)
  32. {
  33. if (unlikely(!wc))
  34. return;
  35. uv_mutex_lock(&aclk_async_lock);
  36. if (unlikely(!wc->host)) {
  37. wc->next = aclk_thread_head;
  38. aclk_thread_head = wc;
  39. }
  40. uv_mutex_unlock(&aclk_async_lock);
  41. return;
  42. }
  43. void aclk_del_worker_thread(struct aclk_database_worker_config *wc)
  44. {
  45. if (unlikely(!wc))
  46. return;
  47. uv_mutex_lock(&aclk_async_lock);
  48. struct aclk_database_worker_config **tmp = &aclk_thread_head;
  49. while (*tmp && (*tmp) != wc)
  50. tmp = &(*tmp)->next;
  51. if (*tmp)
  52. *tmp = wc->next;
  53. uv_mutex_unlock(&aclk_async_lock);
  54. return;
  55. }
  56. int aclk_worker_thread_exists(char *guid)
  57. {
  58. int rc = 0;
  59. uv_mutex_lock(&aclk_async_lock);
  60. struct aclk_database_worker_config *tmp = aclk_thread_head;
  61. while (tmp && !rc) {
  62. rc = strcmp(tmp->uuid_str, guid) == 0;
  63. tmp = tmp->next;
  64. }
  65. uv_mutex_unlock(&aclk_async_lock);
  66. return rc;
  67. }
  68. void aclk_database_init_cmd_queue(struct aclk_database_worker_config *wc)
  69. {
  70. wc->cmd_queue.head = wc->cmd_queue.tail = 0;
  71. wc->queue_size = 0;
  72. fatal_assert(0 == uv_cond_init(&wc->cmd_cond));
  73. fatal_assert(0 == uv_mutex_init(&wc->cmd_mutex));
  74. }
  75. int aclk_database_enq_cmd_noblock(struct aclk_database_worker_config *wc, struct aclk_database_cmd *cmd)
  76. {
  77. unsigned queue_size;
  78. /* wait for free space in queue */
  79. uv_mutex_lock(&wc->cmd_mutex);
  80. if ((queue_size = wc->queue_size) == ACLK_DATABASE_CMD_Q_MAX_SIZE) {
  81. uv_mutex_unlock(&wc->cmd_mutex);
  82. return 1;
  83. }
  84. fatal_assert(queue_size < ACLK_DATABASE_CMD_Q_MAX_SIZE);
  85. /* enqueue command */
  86. wc->cmd_queue.cmd_array[wc->cmd_queue.tail] = *cmd;
  87. wc->cmd_queue.tail = wc->cmd_queue.tail != ACLK_DATABASE_CMD_Q_MAX_SIZE - 1 ?
  88. wc->cmd_queue.tail + 1 : 0;
  89. wc->queue_size = queue_size + 1;
  90. uv_mutex_unlock(&wc->cmd_mutex);
  91. return 0;
  92. }
  93. void aclk_database_enq_cmd(struct aclk_database_worker_config *wc, struct aclk_database_cmd *cmd)
  94. {
  95. unsigned queue_size;
  96. /* wait for free space in queue */
  97. uv_mutex_lock(&wc->cmd_mutex);
  98. while ((queue_size = wc->queue_size) == ACLK_DATABASE_CMD_Q_MAX_SIZE) {
  99. uv_cond_wait(&wc->cmd_cond, &wc->cmd_mutex);
  100. }
  101. fatal_assert(queue_size < ACLK_DATABASE_CMD_Q_MAX_SIZE);
  102. /* enqueue command */
  103. wc->cmd_queue.cmd_array[wc->cmd_queue.tail] = *cmd;
  104. wc->cmd_queue.tail = wc->cmd_queue.tail != ACLK_DATABASE_CMD_Q_MAX_SIZE - 1 ?
  105. wc->cmd_queue.tail + 1 : 0;
  106. wc->queue_size = queue_size + 1;
  107. uv_mutex_unlock(&wc->cmd_mutex);
  108. /* wake up event loop */
  109. int rc = uv_async_send(&wc->async);
  110. if (unlikely(rc))
  111. debug(D_ACLK_SYNC, "Failed to wake up event loop");
  112. }
  113. struct aclk_database_cmd aclk_database_deq_cmd(struct aclk_database_worker_config* wc)
  114. {
  115. struct aclk_database_cmd ret;
  116. unsigned queue_size;
  117. uv_mutex_lock(&wc->cmd_mutex);
  118. queue_size = wc->queue_size;
  119. if (queue_size == 0) {
  120. memset(&ret, 0, sizeof(ret));
  121. ret.opcode = ACLK_DATABASE_NOOP;
  122. ret.completion = NULL;
  123. } else {
  124. /* dequeue command */
  125. ret = wc->cmd_queue.cmd_array[wc->cmd_queue.head];
  126. if (queue_size == 1) {
  127. wc->cmd_queue.head = wc->cmd_queue.tail = 0;
  128. } else {
  129. wc->cmd_queue.head = wc->cmd_queue.head != ACLK_DATABASE_CMD_Q_MAX_SIZE - 1 ?
  130. wc->cmd_queue.head + 1 : 0;
  131. }
  132. wc->queue_size = queue_size - 1;
  133. /* wake up producers */
  134. uv_cond_signal(&wc->cmd_cond);
  135. }
  136. uv_mutex_unlock(&wc->cmd_mutex);
  137. return ret;
  138. }
  139. int aclk_worker_enq_cmd(char *node_id, struct aclk_database_cmd *cmd)
  140. {
  141. if (unlikely(!node_id || !cmd))
  142. return 0;
  143. uv_mutex_lock(&aclk_async_lock);
  144. struct aclk_database_worker_config *wc = aclk_thread_head;
  145. while (wc) {
  146. if (!strcmp(wc->node_id, node_id))
  147. break;
  148. wc = wc->next;
  149. }
  150. uv_mutex_unlock(&aclk_async_lock);
  151. if (wc)
  152. aclk_database_enq_cmd(wc, cmd);
  153. return (wc == NULL);
  154. }
  155. int aclk_start_sync_thread(void *data, int argc, char **argv, char **column)
  156. {
  157. char uuid_str[GUID_LEN + 1];
  158. UNUSED(data);
  159. UNUSED(argc);
  160. UNUSED(column);
  161. uuid_unparse_lower(*((uuid_t *) argv[0]), uuid_str);
  162. if (rrdhost_find_by_guid(uuid_str, 0) == localhost)
  163. return 0;
  164. sql_create_aclk_table(NULL, (uuid_t *) argv[0], (uuid_t *) argv[1]);
  165. return 0;
  166. }
  167. void sql_aclk_sync_init(void)
  168. {
  169. #ifdef ENABLE_NEW_CLOUD_PROTOCOL
  170. char *err_msg = NULL;
  171. int rc;
  172. if (unlikely(!db_meta)) {
  173. if (default_rrd_memory_mode != RRD_MEMORY_MODE_DBENGINE) {
  174. return;
  175. }
  176. error_report("Database has not been initialized");
  177. return;
  178. }
  179. info("SQLite aclk sync initialization");
  180. for (int i = 0; aclk_sync_config[i]; i++) {
  181. debug(D_ACLK_SYNC, "Executing %s", aclk_sync_config[i]);
  182. rc = sqlite3_exec(db_meta, aclk_sync_config[i], 0, 0, &err_msg);
  183. if (rc != SQLITE_OK) {
  184. error_report("SQLite error aclk sync initialization setup, rc = %d (%s)", rc, err_msg);
  185. error_report("SQLite failed statement %s", aclk_sync_config[i]);
  186. sqlite3_free(err_msg);
  187. return;
  188. }
  189. }
  190. info("SQLite aclk sync initialization completed");
  191. fatal_assert(0 == uv_mutex_init(&aclk_async_lock));
  192. rc = sqlite3_exec(db_meta, "SELECT ni.host_id, ni.node_id FROM host h, node_instance ni WHERE "
  193. "h.host_id = ni.host_id AND ni.node_id IS NOT NULL;", aclk_start_sync_thread, NULL, NULL);
  194. #endif
  195. return;
  196. }
  197. static void async_cb(uv_async_t *handle)
  198. {
  199. uv_stop(handle->loop);
  200. uv_update_time(handle->loop);
  201. debug(D_ACLK_SYNC, "%s called, active=%d.", __func__, uv_is_active((uv_handle_t *)handle));
  202. }
  203. #define TIMER_PERIOD_MS (1000)
  204. static void timer_cb(uv_timer_t* handle)
  205. {
  206. uv_stop(handle->loop);
  207. uv_update_time(handle->loop);
  208. #ifdef ENABLE_NEW_CLOUD_PROTOCOL
  209. struct aclk_database_worker_config *wc = handle->data;
  210. struct aclk_database_cmd cmd;
  211. memset(&cmd, 0, sizeof(cmd));
  212. cmd.opcode = ACLK_DATABASE_TIMER;
  213. aclk_database_enq_cmd_noblock(wc, &cmd);
  214. time_t now = now_realtime_sec();
  215. if (wc->cleanup_after && wc->cleanup_after < now) {
  216. cmd.opcode = ACLK_DATABASE_CLEANUP;
  217. if (!aclk_database_enq_cmd_noblock(wc, &cmd))
  218. wc->cleanup_after += ACLK_DATABASE_CLEANUP_INTERVAL;
  219. }
  220. if (aclk_use_new_cloud_arch && aclk_connected) {
  221. if (wc->rotation_after && wc->rotation_after < now) {
  222. cmd.opcode = ACLK_DATABASE_NODE_INFO;
  223. aclk_database_enq_cmd_noblock(wc, &cmd);
  224. cmd.opcode = ACLK_DATABASE_UPD_RETENTION;
  225. if (!aclk_database_enq_cmd_noblock(wc, &cmd))
  226. wc->rotation_after += ACLK_DATABASE_ROTATION_INTERVAL;
  227. }
  228. if (wc->chart_updates && !wc->chart_pending && wc->chart_payload_count) {
  229. cmd.opcode = ACLK_DATABASE_PUSH_CHART;
  230. cmd.count = ACLK_MAX_CHART_BATCH;
  231. cmd.param1 = ACLK_MAX_CHART_BATCH_COUNT;
  232. if (!aclk_database_enq_cmd_noblock(wc, &cmd)) {
  233. if (wc->retry_count)
  234. info("Queued chart/dimension payload command %s, retry count = %u", wc->host_guid, wc->retry_count);
  235. wc->chart_pending = 1;
  236. wc->retry_count = 0;
  237. } else {
  238. wc->retry_count++;
  239. if (wc->retry_count % 100 == 0)
  240. error_report("Failed to queue chart/dimension payload command %s, retry count = %u",
  241. wc->host_guid,
  242. wc->retry_count);
  243. }
  244. }
  245. if (wc->alert_updates) {
  246. cmd.opcode = ACLK_DATABASE_PUSH_ALERT;
  247. cmd.count = ACLK_MAX_ALERT_UPDATES;
  248. aclk_database_enq_cmd_noblock(wc, &cmd);
  249. }
  250. }
  251. #endif
  252. }
  253. #define MAX_CMD_BATCH_SIZE (256)
  254. void aclk_database_worker(void *arg)
  255. {
  256. struct aclk_database_worker_config *wc = arg;
  257. uv_loop_t *loop;
  258. int ret;
  259. enum aclk_database_opcode opcode;
  260. uv_timer_t timer_req;
  261. struct aclk_database_cmd cmd;
  262. unsigned cmd_batch_size;
  263. aclk_database_init_cmd_queue(wc);
  264. char threadname[NETDATA_THREAD_NAME_MAX+1];
  265. if (wc->host)
  266. snprintfz(threadname, NETDATA_THREAD_NAME_MAX, "AS_%s", wc->host->hostname);
  267. else {
  268. snprintfz(threadname, NETDATA_THREAD_NAME_MAX, "AS_%s", wc->uuid_str);
  269. threadname[11] = '\0';
  270. }
  271. uv_thread_set_name_np(wc->thread, threadname);
  272. loop = wc->loop = mallocz(sizeof(uv_loop_t));
  273. ret = uv_loop_init(loop);
  274. if (ret) {
  275. error("uv_loop_init(): %s", uv_strerror(ret));
  276. goto error_after_loop_init;
  277. }
  278. loop->data = wc;
  279. ret = uv_async_init(wc->loop, &wc->async, async_cb);
  280. if (ret) {
  281. error("uv_async_init(): %s", uv_strerror(ret));
  282. goto error_after_async_init;
  283. }
  284. wc->async.data = wc;
  285. ret = uv_timer_init(loop, &timer_req);
  286. if (ret) {
  287. error("uv_timer_init(): %s", uv_strerror(ret));
  288. goto error_after_timer_init;
  289. }
  290. timer_req.data = wc;
  291. fatal_assert(0 == uv_timer_start(&timer_req, timer_cb, TIMER_PERIOD_MS, TIMER_PERIOD_MS));
  292. wc->retry_count = 0;
  293. wc->node_info_send = (wc->host && !localhost);
  294. aclk_add_worker_thread(wc);
  295. info("Starting ACLK sync thread for host %s -- scratch area %lu bytes", wc->host_guid, sizeof(*wc));
  296. memset(&cmd, 0, sizeof(cmd));
  297. #ifdef ENABLE_NEW_CLOUD_PROTOCOL
  298. sql_get_last_chart_sequence(wc);
  299. wc->chart_payload_count = sql_get_pending_count(wc);
  300. if (!wc->chart_payload_count)
  301. info("%s: No pending charts and dimensions detected during startup", wc->host_guid);
  302. #endif
  303. wc->chart_updates = 0;
  304. wc->startup_time = now_realtime_sec();
  305. wc->cleanup_after = wc->startup_time + ACLK_DATABASE_CLEANUP_FIRST;
  306. wc->rotation_after = wc->startup_time + ACLK_DATABASE_ROTATION_DELAY;
  307. wc->alert_updates = 0;
  308. debug(D_ACLK_SYNC,"Node %s reports pending message count = %u", wc->node_id, wc->chart_payload_count);
  309. while (likely(!netdata_exit)) {
  310. uv_run(loop, UV_RUN_DEFAULT);
  311. /* wait for commands */
  312. cmd_batch_size = 0;
  313. do {
  314. if (unlikely(cmd_batch_size >= MAX_CMD_BATCH_SIZE))
  315. break;
  316. cmd = aclk_database_deq_cmd(wc);
  317. if (netdata_exit)
  318. break;
  319. opcode = cmd.opcode;
  320. ++cmd_batch_size;
  321. switch (opcode) {
  322. case ACLK_DATABASE_NOOP:
  323. /* the command queue was empty, do nothing */
  324. break;
  325. // MAINTENANCE
  326. case ACLK_DATABASE_CLEANUP:
  327. debug(D_ACLK_SYNC, "Database cleanup for %s", wc->host_guid);
  328. sql_maint_aclk_sync_database(wc, cmd);
  329. if (wc->host == localhost)
  330. sql_check_aclk_table_list(wc);
  331. break;
  332. case ACLK_DATABASE_DELETE_HOST:
  333. debug(D_ACLK_SYNC,"Cleaning ACLK tables for %s", (char *) cmd.data);
  334. sql_delete_aclk_table_list(wc, cmd);
  335. break;
  336. // CHART / DIMENSION OPERATIONS
  337. #ifdef ENABLE_NEW_CLOUD_PROTOCOL
  338. case ACLK_DATABASE_ADD_CHART:
  339. debug(D_ACLK_SYNC, "Adding chart event for %s", wc->host_guid);
  340. aclk_add_chart_event(wc, cmd);
  341. break;
  342. case ACLK_DATABASE_ADD_DIMENSION:
  343. debug(D_ACLK_SYNC, "Adding dimension event for %s", wc->host_guid);
  344. aclk_add_dimension_event(wc, cmd);
  345. break;
  346. case ACLK_DATABASE_PUSH_CHART:
  347. debug(D_ACLK_SYNC, "Pushing chart info to the cloud for node %s", wc->host_guid);
  348. aclk_send_chart_event(wc, cmd);
  349. break;
  350. case ACLK_DATABASE_PUSH_CHART_CONFIG:
  351. debug(D_ACLK_SYNC, "Pushing chart config info to the cloud for node %s", wc->host_guid);
  352. aclk_send_chart_config(wc, cmd);
  353. break;
  354. case ACLK_DATABASE_CHART_ACK:
  355. debug(D_ACLK_SYNC, "ACK chart SEQ for %s to %"PRIu64, wc->uuid_str, (uint64_t) cmd.param1);
  356. aclk_receive_chart_ack(wc, cmd);
  357. break;
  358. case ACLK_DATABASE_RESET_CHART:
  359. debug(D_ACLK_SYNC, "RESET chart SEQ for %s to %"PRIu64, wc->uuid_str, (uint64_t) cmd.param1);
  360. aclk_receive_chart_reset(wc, cmd);
  361. break;
  362. #endif
  363. // ALERTS
  364. case ACLK_DATABASE_ADD_ALERT:
  365. debug(D_ACLK_SYNC,"Adding alert event for %s", wc->host_guid);
  366. aclk_add_alert_event(wc, cmd);
  367. break;
  368. case ACLK_DATABASE_PUSH_ALERT_CONFIG:
  369. debug(D_ACLK_SYNC,"Pushing chart config info to the cloud for %s", wc->host_guid);
  370. aclk_push_alert_config_event(wc, cmd);
  371. break;
  372. case ACLK_DATABASE_PUSH_ALERT:
  373. debug(D_ACLK_SYNC, "Pushing alert info to the cloud for %s", wc->host_guid);
  374. aclk_push_alert_event(wc, cmd);
  375. break;
  376. case ACLK_DATABASE_ALARM_HEALTH_LOG:
  377. debug(D_ACLK_SYNC, "Pushing alarm health log to the cloud for %s", wc->host_guid);
  378. aclk_push_alarm_health_log(wc, cmd);
  379. break;
  380. case ACLK_DATABASE_PUSH_ALERT_SNAPSHOT:
  381. debug(D_ACLK_SYNC, "Pushing alert snapshot to the cloud for node %s", wc->host_guid);
  382. aclk_push_alert_snapshot_event(wc, cmd);
  383. break;
  384. case ACLK_DATABASE_QUEUE_REMOVED_ALERTS:
  385. debug(D_ACLK_SYNC, "Queueing removed alerts for node %s", wc->host_guid);
  386. sql_process_queue_removed_alerts_to_aclk(wc, cmd);
  387. break;
  388. // NODE OPERATIONS
  389. case ACLK_DATABASE_NODE_INFO:
  390. debug(D_ACLK_SYNC,"Sending node info for %s", wc->uuid_str);
  391. sql_build_node_info(wc, cmd);
  392. break;
  393. #ifdef ENABLE_NEW_CLOUD_PROTOCOL
  394. case ACLK_DATABASE_DIM_DELETION:
  395. debug(D_ACLK_SYNC,"Sending dimension deletion information %s", wc->uuid_str);
  396. aclk_process_dimension_deletion(wc, cmd);
  397. break;
  398. case ACLK_DATABASE_UPD_RETENTION:
  399. debug(D_ACLK_SYNC,"Sending retention info for %s", wc->uuid_str);
  400. aclk_update_retention(wc, cmd);
  401. aclk_process_dimension_deletion(wc, cmd);
  402. break;
  403. #endif
  404. // NODE_INSTANCE DETECTION
  405. case ACLK_DATABASE_TIMER:
  406. if (unlikely(localhost && !wc->host)) {
  407. if (claimed()) {
  408. wc->host = rrdhost_find_by_guid(wc->host_guid, 0);
  409. if (wc->host) {
  410. info("HOST %s (%s) detected as active", wc->host->hostname, wc->host_guid);
  411. snprintfz(threadname, NETDATA_THREAD_NAME_MAX, "AS_%s", wc->host->hostname);
  412. uv_thread_set_name_np(wc->thread, threadname);
  413. wc->host->dbsync_worker = wc;
  414. aclk_del_worker_thread(wc);
  415. wc->node_info_send = 1;
  416. }
  417. }
  418. }
  419. if (wc->node_info_send && wc->host && localhost && claimed() && aclk_connected) {
  420. cmd.opcode = ACLK_DATABASE_NODE_INFO;
  421. cmd.completion = NULL;
  422. wc->node_info_send = aclk_database_enq_cmd_noblock(wc, &cmd);
  423. }
  424. break;
  425. default:
  426. debug(D_ACLK_SYNC, "%s: default.", __func__);
  427. break;
  428. }
  429. if (cmd.completion)
  430. aclk_complete(cmd.completion);
  431. } while (opcode != ACLK_DATABASE_NOOP);
  432. }
  433. if (!uv_timer_stop(&timer_req))
  434. uv_close((uv_handle_t *)&timer_req, NULL);
  435. /* cleanup operations of the event loop */
  436. info("Shutting down ACLK sync event loop.");
  437. /*
  438. * uv_async_send after uv_close does not seem to crash in linux at the moment,
  439. * it is however undocumented behaviour we need to be aware if this becomes
  440. * an issue in the future.
  441. */
  442. uv_close((uv_handle_t *)&wc->async, NULL);
  443. uv_run(loop, UV_RUN_DEFAULT);
  444. info("Shutting down ACLK sync event loop complete.");
  445. /* TODO: don't let the API block by waiting to enqueue commands */
  446. uv_cond_destroy(&wc->cmd_cond);
  447. /* uv_mutex_destroy(&wc->cmd_mutex); */
  448. //fatal_assert(0 == uv_loop_close(loop));
  449. int rc;
  450. do {
  451. rc = uv_loop_close(loop);
  452. } while (rc != UV_EBUSY);
  453. freez(loop);
  454. rrd_wrlock();
  455. if (likely(wc->host))
  456. wc->host->dbsync_worker = NULL;
  457. freez(wc);
  458. rrd_unlock();
  459. return;
  460. error_after_timer_init:
  461. uv_close((uv_handle_t *)&wc->async, NULL);
  462. error_after_async_init:
  463. fatal_assert(0 == uv_loop_close(loop));
  464. error_after_loop_init:
  465. freez(loop);
  466. }
  467. // -------------------------------------------------------------
  468. void sql_create_aclk_table(RRDHOST *host, uuid_t *host_uuid, uuid_t *node_id)
  469. {
  470. #ifdef ENABLE_ACLK
  471. char uuid_str[GUID_LEN + 1];
  472. char host_guid[GUID_LEN + 1];
  473. uuid_unparse_lower_fix(host_uuid, uuid_str);
  474. if (aclk_worker_thread_exists(uuid_str))
  475. return;
  476. uuid_unparse_lower(*host_uuid, host_guid);
  477. BUFFER *sql = buffer_create(ACLK_SYNC_QUERY_SIZE);
  478. buffer_sprintf(sql, TABLE_ACLK_CHART, uuid_str);
  479. db_execute(buffer_tostring(sql));
  480. buffer_flush(sql);
  481. buffer_sprintf(sql, TABLE_ACLK_CHART_PAYLOAD, uuid_str);
  482. db_execute(buffer_tostring(sql));
  483. buffer_flush(sql);
  484. buffer_sprintf(sql, TABLE_ACLK_CHART_LATEST, uuid_str);
  485. db_execute(buffer_tostring(sql));
  486. buffer_flush(sql);
  487. buffer_sprintf(sql, INDEX_ACLK_CHART, uuid_str, uuid_str);
  488. db_execute(buffer_tostring(sql));
  489. buffer_flush(sql);
  490. buffer_sprintf(sql, INDEX_ACLK_CHART_LATEST, uuid_str, uuid_str);
  491. db_execute(buffer_tostring(sql));
  492. buffer_flush(sql);
  493. buffer_sprintf(sql, TRIGGER_ACLK_CHART_PAYLOAD, uuid_str, uuid_str, uuid_str);
  494. db_execute(buffer_tostring(sql));
  495. buffer_flush(sql);
  496. buffer_sprintf(sql, TABLE_ACLK_ALERT, uuid_str, uuid_str, uuid_str);
  497. db_execute(buffer_tostring(sql));
  498. buffer_flush(sql);
  499. buffer_sprintf(sql, INDEX_ACLK_ALERT, uuid_str, uuid_str);
  500. db_execute(buffer_tostring(sql));
  501. buffer_free(sql);
  502. if (likely(host) && unlikely(host->dbsync_worker))
  503. return;
  504. struct aclk_database_worker_config *wc = callocz(1, sizeof(struct aclk_database_worker_config));
  505. if (likely(host))
  506. host->dbsync_worker = (void *) wc;
  507. wc->host = host;
  508. strcpy(wc->uuid_str, uuid_str);
  509. strcpy(wc->host_guid, host_guid);
  510. if (node_id && !uuid_is_null(*node_id))
  511. uuid_unparse_lower(*node_id, wc->node_id);
  512. fatal_assert(0 == uv_thread_create(&(wc->thread), aclk_database_worker, wc));
  513. #else
  514. UNUSED(host);
  515. UNUSED(host_uuid);
  516. UNUSED(node_id);
  517. #endif
  518. return;
  519. }
  520. void sql_maint_aclk_sync_database(struct aclk_database_worker_config *wc, struct aclk_database_cmd cmd)
  521. {
  522. UNUSED(cmd);
  523. debug(D_ACLK, "Checking database for %s", wc->host_guid);
  524. BUFFER *sql = buffer_create(ACLK_SYNC_QUERY_SIZE);
  525. buffer_sprintf(sql,"DELETE FROM aclk_chart_%s WHERE date_submitted IS NOT NULL AND "
  526. "date_updated < strftime('%%s','now','-%d seconds');", wc->uuid_str, ACLK_DELETE_ACK_INTERNAL);
  527. db_execute(buffer_tostring(sql));
  528. buffer_flush(sql);
  529. buffer_sprintf(sql,"DELETE FROM aclk_chart_payload_%s WHERE unique_id NOT IN "
  530. "(SELECT unique_id FROM aclk_chart_%s) AND unique_id NOT IN (SELECT unique_id FROM aclk_chart_latest_%s);",
  531. wc->uuid_str, wc->uuid_str, wc->uuid_str);
  532. db_execute(buffer_tostring(sql));
  533. buffer_free(sql);
  534. return;
  535. }
  536. #define SQL_SELECT_HOST_BY_UUID "SELECT host_id FROM host WHERE host_id = @host_id;"
  537. static int is_host_available(uuid_t *host_id)
  538. {
  539. sqlite3_stmt *res = NULL;
  540. int rc;
  541. if (unlikely(!db_meta)) {
  542. if (default_rrd_memory_mode == RRD_MEMORY_MODE_DBENGINE)
  543. error_report("Database has not been initialized");
  544. return 1;
  545. }
  546. rc = sqlite3_prepare_v2(db_meta, SQL_SELECT_HOST_BY_UUID, -1, &res, 0);
  547. if (unlikely(rc != SQLITE_OK)) {
  548. error_report("Failed to prepare statement to select node instance information for a node");
  549. return 1;
  550. }
  551. rc = sqlite3_bind_blob(res, 1, host_id, sizeof(*host_id), SQLITE_STATIC);
  552. if (unlikely(rc != SQLITE_OK)) {
  553. error_report("Failed to bind host_id parameter to select node instance information");
  554. goto failed;
  555. }
  556. rc = sqlite3_step(res);
  557. failed:
  558. if (unlikely(sqlite3_finalize(res) != SQLITE_OK))
  559. error_report("Failed to finalize the prepared statement when checking host existence");
  560. return (rc == SQLITE_ROW);
  561. }
  562. // OPCODE: ACLK_DATABASE_DELETE_HOST
  563. void sql_delete_aclk_table_list(struct aclk_database_worker_config *wc, struct aclk_database_cmd cmd)
  564. {
  565. UNUSED(wc);
  566. char uuid_str[GUID_LEN + 1];
  567. char host_str[GUID_LEN + 1];
  568. int rc;
  569. uuid_t host_uuid;
  570. char *host_guid = (char *)cmd.data;
  571. if (unlikely(!host_guid))
  572. return;
  573. rc = uuid_parse(host_guid, host_uuid);
  574. freez(host_guid);
  575. if (rc)
  576. return;
  577. uuid_unparse_lower(host_uuid, host_str);
  578. uuid_unparse_lower_fix(&host_uuid, uuid_str);
  579. debug(D_ACLK_SYNC, "Checking if I should delete aclk tables for node %s", host_str);
  580. if (is_host_available(&host_uuid)) {
  581. debug(D_ACLK_SYNC, "Host %s exists, not deleting aclk sync tables", host_str);
  582. return;
  583. }
  584. debug(D_ACLK_SYNC, "Host %s does NOT exist, can delete aclk sync tables", host_str);
  585. sqlite3_stmt *res = NULL;
  586. BUFFER *sql = buffer_create(ACLK_SYNC_QUERY_SIZE);
  587. buffer_sprintf(sql,"SELECT 'drop '||type||' IF EXISTS '||name||';' FROM sqlite_schema " \
  588. "WHERE name LIKE 'aclk_%%_%s' AND type IN ('table', 'trigger', 'index');", uuid_str);
  589. rc = sqlite3_prepare_v2(db_meta, buffer_tostring(sql), -1, &res, 0);
  590. if (rc != SQLITE_OK) {
  591. error_report("Failed to prepare statement to clean up aclk tables");
  592. goto fail;
  593. }
  594. buffer_flush(sql);
  595. while (sqlite3_step(res) == SQLITE_ROW)
  596. buffer_strcat(sql, (char *) sqlite3_column_text(res, 0));
  597. rc = sqlite3_finalize(res);
  598. if (unlikely(rc != SQLITE_OK))
  599. error_report("Failed to finalize statement to clean up aclk tables, rc = %d", rc);
  600. db_execute(buffer_tostring(sql));
  601. fail:
  602. buffer_free(sql);
  603. return;
  604. }
  605. static int sql_check_aclk_table(void *data, int argc, char **argv, char **column)
  606. {
  607. struct aclk_database_worker_config *wc = data;
  608. UNUSED(argc);
  609. UNUSED(column);
  610. debug(D_ACLK_SYNC,"Scheduling aclk sync table check for node %s", (char *) argv[0]);
  611. struct aclk_database_cmd cmd;
  612. memset(&cmd, 0, sizeof(cmd));
  613. cmd.opcode = ACLK_DATABASE_DELETE_HOST;
  614. cmd.data = strdupz((char *) argv[0]);
  615. aclk_database_enq_cmd_noblock(wc, &cmd);
  616. return 0;
  617. }
  618. #define SQL_SELECT_ACLK_ACTIVE_LIST "SELECT REPLACE(SUBSTR(name,19),'_','-') FROM sqlite_schema " \
  619. "WHERE name LIKE 'aclk_chart_latest_%' AND type IN ('table');"
  620. void sql_check_aclk_table_list(struct aclk_database_worker_config *wc)
  621. {
  622. char *err_msg = NULL;
  623. debug(D_ACLK_SYNC,"Cleaning tables for nodes that do not exist");
  624. int rc = sqlite3_exec(db_meta, SQL_SELECT_ACLK_ACTIVE_LIST, sql_check_aclk_table, (void *) wc, &err_msg);
  625. if (rc != SQLITE_OK) {
  626. error_report("Query failed when trying to check for obsolete ACLK sync tables, %s", err_msg);
  627. sqlite3_free(err_msg);
  628. }
  629. db_execute("DELETE FROM dimension_delete WHERE host_id NOT IN (SELECT host_id FROM host) "
  630. " OR strftime('%s') - date_created > 604800;");
  631. return;
  632. }
  633. void aclk_data_rotated(void)
  634. {
  635. #ifdef ENABLE_NEW_CLOUD_PROTOCOL
  636. if (!aclk_use_new_cloud_arch || !aclk_connected)
  637. return;
  638. time_t next_rotation_time = now_realtime_sec()+ACLK_DATABASE_ROTATION_DELAY;
  639. rrd_wrlock();
  640. RRDHOST *this_host = localhost;
  641. while (this_host) {
  642. struct aclk_database_worker_config *wc = this_host->dbsync_worker;
  643. if (wc)
  644. wc->rotation_after = next_rotation_time;
  645. this_host = this_host->next;
  646. }
  647. rrd_unlock();
  648. struct aclk_database_worker_config *tmp = aclk_thread_head;
  649. uv_mutex_lock(&aclk_async_lock);
  650. while (tmp) {
  651. tmp->rotation_after = next_rotation_time;
  652. tmp = tmp->next;
  653. }
  654. uv_mutex_unlock(&aclk_async_lock);
  655. #endif
  656. return;
  657. }