testcpg.c 12 KB

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101102103104105106107108109110111112113114115116117118119120121122123124125126127128129130131132133134135136137138139140141142143144145146147148149150151152153154155156157158159160161162163164165166167168169170171172173174175176177178179180181182183184185186187188189190191192193194195196197198199200201202203204205206207208209210211212213214215216217218219220221222223224225226227228229230231232233234235236237238239240241242243244245246247248249250251252253254255256257258259260261262263264265266267268269270271272273274275276277278279280281282283284285286287288289290291292293294295296297298299300301302303304305306307308309310311312313314315316317318319320321322323324325326327328329330331332333334335336337338339340341342343344345346347348349350351352353354355356357358359360361362363364365366367368369370371372373374375376377378379380381382383384385386387388389390391392393394395396397398399400401402403404405406407408409410411412413414415416417418419420421422423424425426427428429430431432433434435436
  1. /*
  2. * Copyright (c) 2006-2009 Red Hat Inc
  3. *
  4. * All rights reserved.
  5. *
  6. * Author: Christine Caulfield <ccaulfie@redhat.com>
  7. *
  8. * This software licensed under BSD license, the text of which follows:
  9. *
  10. * Redistribution and use in source and binary forms, with or without
  11. * modification, are permitted provided that the following conditions are met:
  12. *
  13. * - Redistributions of source code must retain the above copyright notice,
  14. * this list of conditions and the following disclaimer.
  15. * - Redistributions in binary form must reproduce the above copyright notice,
  16. * this list of conditions and the following disclaimer in the documentation
  17. * and/or other materials provided with the distribution.
  18. * - Neither the name of the MontaVista Software, Inc. nor the names of its
  19. * contributors may be used to endorse or promote products derived from this
  20. * software without specific prior written permission.
  21. *
  22. * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
  23. * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
  24. * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
  25. * ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE
  26. * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
  27. * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
  28. * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
  29. * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
  30. * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
  31. * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF
  32. * THE POSSIBILITY OF SUCH DAMAGE.
  33. */
  34. #include <config.h>
  35. #include <inttypes.h>
  36. #include <stdio.h>
  37. #include <stdlib.h>
  38. #include <errno.h>
  39. #include <unistd.h>
  40. #include <string.h>
  41. #include <sys/types.h>
  42. #include <sys/socket.h>
  43. #include <sys/select.h>
  44. #include <sys/un.h>
  45. #include <netinet/in.h>
  46. #include <arpa/inet.h>
  47. #include <time.h>
  48. #include <sys/time.h>
  49. #include <assert.h>
  50. #include <limits.h>
  51. #include <corosync/corotypes.h>
  52. #include <corosync/cpg.h>
  53. #include <corosync/swab.h>
  54. #ifdef QBLOG
  55. #include <qb/qblog.h>
  56. #endif
  57. #ifndef HOST_NAME_MAX
  58. #define HOST_NAME_MAX _POSIX_HOST_NAME_MAX
  59. #endif
  60. static int quit = 0;
  61. static int show_ip = 0;
  62. static int restart = 0;
  63. static uint32_t nodeidStart = 0;
  64. static void print_localnodeid(cpg_handle_t handle);
  65. static void print_cpgname (const struct cpg_name *name)
  66. {
  67. unsigned int i;
  68. for (i = 0; i < name->length; i++) {
  69. printf ("%c", name->value[i]);
  70. }
  71. }
  72. static char * node_pid_format(unsigned int nodeid, unsigned int pid) {
  73. static char buffer[100];
  74. if (show_ip) {
  75. struct in_addr saddr;
  76. #if __BYTE_ORDER == __BIG_ENDIAN
  77. saddr.s_addr = swab32(nodeid);
  78. #else
  79. saddr.s_addr = nodeid;
  80. #endif
  81. sprintf(buffer, "node/pid %s/%d", inet_ntoa(saddr),pid);
  82. }
  83. else {
  84. sprintf(buffer, "node/pid %d/%d", nodeid, pid);
  85. }
  86. return buffer;
  87. }
  88. static void
  89. print_time(void)
  90. {
  91. #define MAXLEN (256)
  92. char buf[MAXLEN];
  93. char hostname[HOST_NAME_MAX];
  94. struct timeval tnow;
  95. time_t t;
  96. size_t len;
  97. char *s = buf;
  98. len = sizeof(hostname);
  99. if(gethostname(hostname, len) == 0) {
  100. char *longName;
  101. hostname[len-1] = '\0';
  102. longName = hostname;
  103. if( (longName = strstr( hostname, "." )) != NULL )
  104. *longName = '\0';
  105. }
  106. strcpy(s, hostname);
  107. s += strlen(hostname);
  108. s += snprintf(s, sizeof(buf)-(s-buf), ":%d", getpid());
  109. t = time(0);
  110. gettimeofday( &tnow, 0 );
  111. s += strftime(s, sizeof(buf)-(s-buf) , " %Y-%m-%d %T", localtime(&t));
  112. s += snprintf(s, sizeof(buf)-(s-buf), ".%03ld", tnow.tv_usec/1000);
  113. assert(s-buf < (int)sizeof(buf));
  114. printf("%s\n", buf);
  115. }
  116. static void DeliverCallback (
  117. cpg_handle_t handle,
  118. const struct cpg_name *groupName,
  119. uint32_t nodeid,
  120. uint32_t pid,
  121. void *msg,
  122. size_t msg_len)
  123. {
  124. print_time();
  125. printf("DeliverCallback: message (len=%lu)from %s: '%s'\n",
  126. (unsigned long int) msg_len, node_pid_format(nodeid, pid),
  127. (const char *)msg);
  128. }
  129. static void ConfchgCallback (
  130. cpg_handle_t handle,
  131. const struct cpg_name *groupName,
  132. const struct cpg_address *member_list, size_t member_list_entries,
  133. const struct cpg_address *left_list, size_t left_list_entries,
  134. const struct cpg_address *joined_list, size_t joined_list_entries)
  135. {
  136. unsigned int i;
  137. int result;
  138. uint32_t nodeid;
  139. print_time();
  140. printf("ConfchgCallback: group '");
  141. print_cpgname(groupName);
  142. printf("'\n");
  143. print_localnodeid(handle);
  144. for (i=0; i<joined_list_entries; i++) {
  145. printf("joined %s reason: %d\n",
  146. node_pid_format(joined_list[i].nodeid, joined_list[i].pid),
  147. joined_list[i].reason);
  148. }
  149. for (i=0; i<left_list_entries; i++) {
  150. printf("left %s reason: %d\n",
  151. node_pid_format(left_list[i].nodeid, left_list[i].pid),
  152. left_list[i].reason);
  153. }
  154. printf("nodes in group now %lu\n",
  155. (unsigned long int) member_list_entries);
  156. for (i=0; i<member_list_entries; i++) {
  157. printf("%s\n",
  158. node_pid_format(member_list[i].nodeid, member_list[i].pid));
  159. }
  160. result = cpg_local_get(handle, &nodeid);
  161. if(result != CS_OK) {
  162. printf("failed to get local nodeid %d\n", result);
  163. nodeid = 0;
  164. }
  165. /* Is it us??
  166. NOTE: in reality we should also check the nodeid */
  167. if (left_list_entries && (pid_t)left_list[0].pid == getpid()) {
  168. printf("We might have left the building pid %d\n", left_list[0].pid);
  169. /* can only use nodeidStart as a reliable check (version <= 1.4.2) */
  170. if(nodeidStart) {
  171. /* report dynamic nature of nodeid returned from local_get */
  172. /* local get of nodeid might change identity from original! */
  173. if(htonl((uint32_t)nodeid) == INADDR_LOOPBACK) {
  174. printf("We probably left the building switched identity? start nodeid %d nodeid %d current nodeid %d pid %d\n", nodeidStart, left_list[0].nodeid, nodeid, left_list[0].pid);
  175. } else if(htonl((uint32_t)left_list[0].nodeid) == INADDR_LOOPBACK) {
  176. printf("We probably left the building started alone? start nodeid %d nodeid %d current nodeid %d pid %d\n", nodeidStart, left_list[0].nodeid, nodeid, left_list[0].pid);
  177. }
  178. /* a possibly reliable way to check is based on original address */
  179. if(left_list[0].nodeid == nodeidStart) {
  180. printf("We have left the building direct match start nodeid %d nodeid %d local get current nodeid %d pid %d\n", nodeidStart, left_list[0].nodeid, nodeid, left_list[0].pid);
  181. // quit = 1;
  182. restart = 1;
  183. } else {
  184. printf("Probably another node with matching pid start nodeid %d nodeid %d current nodeid %d pid %d\n", nodeidStart, left_list[0].nodeid, nodeid, left_list[0].pid);
  185. }
  186. }
  187. }
  188. }
  189. static void TotemConfchgCallback (
  190. cpg_handle_t handle,
  191. struct cpg_ring_id ring_id,
  192. uint32_t member_list_entries,
  193. const uint32_t *member_list)
  194. {
  195. unsigned int i;
  196. printf("\n");
  197. print_time();
  198. printf ("TotemConfchgCallback: ringid (%u.%"PRIu64")\n",
  199. ring_id.nodeid, ring_id.seq);
  200. printf("active processors %lu: ",
  201. (unsigned long int) member_list_entries);
  202. for (i=0; i<member_list_entries; i++) {
  203. printf("%d ", member_list[i]);
  204. }
  205. printf ("\n");
  206. }
  207. static cpg_model_v1_data_t model_data = {
  208. .cpg_deliver_fn = DeliverCallback,
  209. .cpg_confchg_fn = ConfchgCallback,
  210. .cpg_totem_confchg_fn = TotemConfchgCallback,
  211. .flags = CPG_MODEL_V1_DELIVER_INITIAL_TOTEM_CONF,
  212. };
  213. static void sigintr_handler (int signum) __attribute__((noreturn));
  214. static void sigintr_handler (int signum) {
  215. exit (0);
  216. }
  217. static struct cpg_name group_name;
  218. #define retrybackoff(counter) { \
  219. counter++; \
  220. printf("Restart operation after %ds\n", counter); \
  221. sleep((unsigned int)counter); \
  222. restart = 1; \
  223. continue; \
  224. }
  225. #define cs_repeat_init(counter, max, code) do { \
  226. code; \
  227. if (result == CS_ERR_TRY_AGAIN || result == CS_ERR_QUEUE_FULL || result == CS_ERR_LIBRARY) { \
  228. counter++; \
  229. printf("Retrying operation after %ds\n", counter); \
  230. sleep((unsigned int)counter); \
  231. } else { \
  232. break; \
  233. } \
  234. } while (counter < max)
  235. #define cs_repeat(counter, max, code) do { \
  236. code; \
  237. if (result == CS_ERR_TRY_AGAIN || result == CS_ERR_QUEUE_FULL) { \
  238. counter++; \
  239. printf("Retrying operation after %ds\n", counter); \
  240. sleep((unsigned int)counter); \
  241. } else { \
  242. break; \
  243. } \
  244. } while (counter < max)
  245. static void print_localnodeid(cpg_handle_t handle)
  246. {
  247. char addrStr[128];
  248. unsigned int retries;
  249. unsigned int nodeid;
  250. struct sockaddr_storage addr;
  251. struct sockaddr_in *v4addr = (struct sockaddr_in *)&addr;
  252. int result;
  253. retries = 0;
  254. cs_repeat(retries, 30, result = cpg_local_get(handle, &nodeid));
  255. if (result != CS_OK) {
  256. printf ("Could not get local node id\n");
  257. } else {
  258. v4addr->sin_addr.s_addr = nodeid;
  259. if(inet_ntop(AF_INET, (const void *)&v4addr->sin_addr.s_addr,
  260. addrStr, (socklen_t)sizeof(addrStr)) == NULL) {
  261. addrStr[0] = 0;
  262. }
  263. printf ("Local node id is %s/%x result %d\n", addrStr, nodeid, result);
  264. }
  265. }
  266. int main (int argc, char *argv[]) {
  267. cpg_handle_t handle;
  268. fd_set read_fds;
  269. int select_fd;
  270. int result;
  271. int retries;
  272. const char *options = "i";
  273. int opt;
  274. unsigned int nodeid;
  275. char *fgets_res;
  276. struct cpg_address member_list[64];
  277. int member_list_entries;
  278. int i;
  279. int recnt;
  280. int doexit;
  281. const char *exitStr = "EXIT";
  282. doexit = 0;
  283. #ifdef QBLOG
  284. qb_log_init("testcpg", LOG_USER, LOG_ERR);
  285. qb_log_ctl(QB_LOG_SYSLOG, QB_LOG_CONF_ENABLED, QB_FALSE);
  286. qb_log_filter_ctl(QB_LOG_STDERR, QB_LOG_FILTER_ADD,
  287. QB_LOG_FILTER_FILE, "*", LOG_TRACE);
  288. qb_log_ctl(QB_LOG_STDERR, QB_LOG_CONF_ENABLED, QB_TRUE);
  289. qb_log_format_set(QB_LOG_STDERR, "[%p] %f %b");
  290. #endif
  291. while ( (opt = getopt(argc, argv, options)) != -1 ) {
  292. switch (opt) {
  293. case 'i':
  294. show_ip = 1;
  295. break;
  296. }
  297. }
  298. if (argc > optind) {
  299. strcpy(group_name.value, argv[optind]);
  300. group_name.length = strlen(argv[optind]);
  301. }
  302. else {
  303. strcpy(group_name.value, "GROUP");
  304. group_name.length = 6;
  305. }
  306. recnt = 0;
  307. printf ("Type %s to finish\n", exitStr);
  308. restart = 1;
  309. do {
  310. if(restart) {
  311. restart = 0;
  312. retries = 0;
  313. cs_repeat_init(retries, 30, result = cpg_model_initialize (&handle, CPG_MODEL_V1, (cpg_model_data_t *)&model_data, NULL));
  314. if (result != CS_OK) {
  315. printf ("Could not initialize Cluster Process Group API instance error %d\n", result);
  316. retrybackoff(recnt);
  317. }
  318. retries = 0;
  319. cs_repeat(retries, 30, result = cpg_local_get(handle, &nodeid));
  320. if (result != CS_OK) {
  321. printf ("Could not get local node id\n");
  322. retrybackoff(recnt);
  323. }
  324. printf ("Local node id is %x\n", nodeid);
  325. nodeidStart = nodeid;
  326. retries = 0;
  327. cs_repeat(retries, 30, result = cpg_join(handle, &group_name));
  328. if (result != CS_OK) {
  329. printf ("Could not join process group, error %d\n", result);
  330. retrybackoff(recnt);
  331. }
  332. retries = 0;
  333. cs_repeat(retries, 30, result = cpg_membership_get (handle, &group_name,
  334. (struct cpg_address *)&member_list, &member_list_entries));
  335. if (result != CS_OK) {
  336. printf ("Could not get current membership list %d\n", result);
  337. retrybackoff(recnt);
  338. }
  339. recnt = 0;
  340. printf ("membership list\n");
  341. for (i = 0; i < member_list_entries; i++) {
  342. printf ("node id %d pid %d\n", member_list[i].nodeid,
  343. member_list[i].pid);
  344. }
  345. FD_ZERO (&read_fds);
  346. cpg_fd_get(handle, &select_fd);
  347. }
  348. FD_SET (select_fd, &read_fds);
  349. FD_SET (STDIN_FILENO, &read_fds);
  350. result = select (select_fd + 1, &read_fds, 0, 0, 0);
  351. if (result == -1) {
  352. perror ("select\n");
  353. }
  354. if (FD_ISSET (STDIN_FILENO, &read_fds)) {
  355. char inbuf[132];
  356. struct iovec iov;
  357. fgets_res = fgets(inbuf, (int)sizeof(inbuf), stdin);
  358. if (fgets_res == NULL) {
  359. doexit = 1;
  360. cpg_leave(handle, &group_name);
  361. }
  362. if (strncmp(inbuf, exitStr, strlen(exitStr)) == 0) {
  363. doexit = 1;
  364. cpg_leave(handle, &group_name);
  365. }
  366. else {
  367. iov.iov_base = inbuf;
  368. iov.iov_len = strlen(inbuf)+1;
  369. cpg_mcast_joined(handle, CPG_TYPE_AGREED, &iov, 1);
  370. }
  371. }
  372. if (FD_ISSET (select_fd, &read_fds)) {
  373. if (cpg_dispatch (handle, CS_DISPATCH_ALL) != CS_OK) {
  374. if(doexit) {
  375. exit(1);
  376. }
  377. restart = 1;
  378. }
  379. }
  380. if(restart) {
  381. if(!doexit) {
  382. result = cpg_finalize (handle);
  383. printf ("Finalize+restart result is %d (should be 1)\n", result);
  384. continue;
  385. }
  386. }
  387. } while (result && !quit && !doexit);
  388. result = cpg_finalize (handle);
  389. printf ("Finalize result is %d (should be 1)\n", result);
  390. return (0);
  391. }