CV1.cpp 69 KB

1234567891011121314151617181920212223242526272829303132333435363738394041424344454647484950515253545556575859606162636465666768697071727374757677787980818283848586878889909192939495969798991001011021031041051061071081091101111121131141151161171181191201211221231241251261271281291301311321331341351361371381391401411421431441451461471481491501511521531541551561571581591601611621631641651661671681691701711721731741751761771781791801811821831841851861871881891901911921931941951961971981992002012022032042052062072082092102112122132142152162172182192202212222232242252262272282292302312322332342352362372382392402412422432442452462472482492502512522532542552562572582592602612622632642652662672682692702712722732742752762772782792802812822832842852862872882892902912922932942952962972982993003013023033043053063073083093103113123133143153163173183193203213223233243253263273283293303313323333343353363373383393403413423433443453463473483493503513523533543553563573583593603613623633643653663673683693703713723733743753763773783793803813823833843853863873883893903913923933943953963973983994004014024034044054064074084094104114124134144154164174184194204214224234244254264274284294304314324334344354364374384394404414424434444454464474484494504514524534544554564574584594604614624634644654664674684694704714724734744754764774784794804814824834844854864874884894904914924934944954964974984995005015025035045055065075085095105115125135145155165175185195205215225235245255265275285295305315325335345355365375385395405415425435445455465475485495505515525535545555565575585595605615625635645655665675685695705715725735745755765775785795805815825835845855865875885895905915925935945955965975985996006016026036046056066076086096106116126136146156166176186196206216226236246256266276286296306316326336346356366376386396406416426436446456466476486496506516526536546556566576586596606616626636646656666676686696706716726736746756766776786796806816826836846856866876886896906916926936946956966976986997007017027037047057067077087097107117127137147157167177187197207217227237247257267277287297307317327337347357367377387397407417427437447457467477487497507517527537547557567577587597607617627637647657667677687697707717727737747757767777787797807817827837847857867877887897907917927937947957967977987998008018028038048058068078088098108118128138148158168178188198208218228238248258268278288298308318328338348358368378388398408418428438448458468478488498508518528538548558568578588598608618628638648658668678688698708718728738748758768778788798808818828838848858868878888898908918928938948958968978988999009019029039049059069079089099109119129139149159169179189199209219229239249259269279289299309319329339349359369379389399409419429439449459469479489499509519529539549559569579589599609619629639649659669679689699709719729739749759769779789799809819829839849859869879889899909919929939949959969979989991000100110021003100410051006100710081009101010111012101310141015101610171018101910201021102210231024102510261027102810291030103110321033103410351036103710381039104010411042104310441045104610471048104910501051105210531054105510561057105810591060106110621063106410651066106710681069107010711072107310741075107610771078107910801081108210831084108510861087108810891090109110921093109410951096109710981099110011011102110311041105110611071108110911101111111211131114111511161117111811191120112111221123112411251126112711281129113011311132113311341135113611371138113911401141114211431144114511461147114811491150115111521153115411551156115711581159116011611162116311641165116611671168116911701171117211731174117511761177117811791180118111821183118411851186118711881189119011911192119311941195119611971198119912001201120212031204120512061207120812091210121112121213121412151216121712181219122012211222122312241225122612271228122912301231123212331234123512361237123812391240124112421243124412451246124712481249125012511252125312541255125612571258125912601261126212631264126512661267126812691270127112721273127412751276127712781279128012811282128312841285128612871288128912901291129212931294129512961297129812991300130113021303130413051306130713081309131013111312131313141315131613171318131913201321132213231324132513261327132813291330133113321333133413351336133713381339134013411342134313441345134613471348134913501351135213531354135513561357135813591360136113621363136413651366136713681369137013711372137313741375137613771378137913801381138213831384138513861387138813891390139113921393139413951396139713981399140014011402140314041405140614071408140914101411141214131414141514161417141814191420142114221423142414251426142714281429143014311432143314341435143614371438143914401441144214431444144514461447144814491450145114521453145414551456145714581459146014611462146314641465146614671468146914701471147214731474147514761477147814791480148114821483148414851486148714881489149014911492149314941495149614971498149915001501150215031504150515061507150815091510151115121513151415151516151715181519152015211522152315241525152615271528152915301531153215331534153515361537153815391540154115421543154415451546154715481549155015511552155315541555155615571558155915601561156215631564156515661567156815691570157115721573157415751576157715781579158015811582158315841585158615871588158915901591159215931594159515961597159815991600160116021603160416051606160716081609161016111612161316141615161616171618161916201621162216231624162516261627162816291630163116321633163416351636163716381639164016411642164316441645164616471648164916501651165216531654165516561657165816591660166116621663166416651666166716681669167016711672167316741675167616771678167916801681168216831684168516861687168816891690169116921693169416951696169716981699170017011702170317041705170617071708170917101711171217131714171517161717171817191720172117221723172417251726172717281729173017311732173317341735173617371738173917401741174217431744174517461747174817491750175117521753175417551756175717581759176017611762176317641765176617671768176917701771177217731774177517761777177817791780178117821783178417851786178717881789179017911792179317941795179617971798179918001801180218031804180518061807180818091810181118121813181418151816181718181819182018211822182318241825182618271828182918301831183218331834183518361837183818391840184118421843184418451846184718481849185018511852185318541855185618571858185918601861186218631864186518661867186818691870187118721873187418751876187718781879188018811882188318841885188618871888188918901891189218931894189518961897189818991900190119021903190419051906190719081909191019111912191319141915191619171918191919201921192219231924192519261927192819291930193119321933193419351936193719381939194019411942194319441945194619471948194919501951195219531954195519561957195819591960196119621963196419651966196719681969197019711972197319741975197619771978197919801981198219831984198519861987198819891990199119921993199419951996199719981999200020012002200320042005200620072008200920102011201220132014201520162017201820192020202120222023202420252026202720282029203020312032203320342035203620372038203920402041204220432044204520462047204820492050205120522053205420552056205720582059206020612062206320642065206620672068206920702071207220732074207520762077207820792080208120822083208420852086208720882089209020912092209320942095209620972098209921002101
  1. /* (c) ZeroTier, Inc.
  2. * See LICENSE.txt in nonfree/
  3. */
  4. #include "CV1.hpp"
  5. #ifdef ZT_CONTROLLER_USE_LIBPQ
  6. #include "../../node/Constants.hpp"
  7. #include "../../node/SHA512.hpp"
  8. #include "../../version.h"
  9. #include "CtlUtil.hpp"
  10. #include "EmbeddedNetworkController.hpp"
  11. #include "Redis.hpp"
  12. #include "opentelemetry/trace/provider.h"
  13. #include <chrono>
  14. #include <climits>
  15. #include <iomanip>
  16. #include <libpq-fe.h>
  17. #include <smeeclient.h>
  18. #include <sstream>
  19. // #define REDIS_TRACE 1
  20. using json = nlohmann::json;
  21. namespace {
  22. static const int DB_MINIMUM_VERSION = 38;
  23. } // anonymous namespace
  24. using namespace ZeroTier;
  25. using Attrs = std::vector<std::pair<std::string, std::string> >;
  26. using Item = std::pair<std::string, Attrs>;
  27. using ItemStream = std::vector<Item>;
  28. CV1::CV1(const Identity& myId, const char* path, int listenPort, RedisConfig* rc)
  29. : DB()
  30. , _pool()
  31. , _myId(myId)
  32. , _myAddress(myId.address())
  33. , _ready(0)
  34. , _connected(1)
  35. , _run(1)
  36. , _waitNoticePrinted(false)
  37. , _listenPort(listenPort)
  38. , _rc(rc)
  39. , _redis(NULL)
  40. , _cluster(NULL)
  41. , _redisMemberStatus(false)
  42. , _smee(NULL)
  43. {
  44. auto provider = opentelemetry::trace::Provider::GetTracerProvider();
  45. auto tracer = provider->GetTracer("cv1");
  46. auto span = tracer->StartSpan("cv1::CV1");
  47. auto scope = tracer->WithActiveSpan(span);
  48. char myAddress[64];
  49. _myAddressStr = myId.address().toString(myAddress);
  50. _connString = std::string(path);
  51. auto f = std::make_shared<PostgresConnFactory>(_connString);
  52. _pool = std::make_shared<ConnectionPool<PostgresConnection> >(15, 5, std::static_pointer_cast<ConnectionFactory>(f));
  53. memset(_ssoPsk, 0, sizeof(_ssoPsk));
  54. char* const ssoPskHex = getenv("ZT_SSO_PSK");
  55. #ifdef ZT_TRACE
  56. fprintf(stderr, "ZT_SSO_PSK: %s\n", ssoPskHex);
  57. #endif
  58. if (ssoPskHex) {
  59. // SECURITY: note that ssoPskHex will always be null-terminated if libc actually
  60. // returns something non-NULL. If the hex encodes something shorter than 48 bytes,
  61. // it will be padded at the end with zeroes. If longer, it'll be truncated.
  62. Utils::unhex(ssoPskHex, _ssoPsk, sizeof(_ssoPsk));
  63. }
  64. const char* redisMemberStatus = getenv("ZT_REDIS_MEMBER_STATUS");
  65. if (redisMemberStatus && (strcmp(redisMemberStatus, "true") == 0)) {
  66. _redisMemberStatus = true;
  67. fprintf(stderr, "Using redis for member status\n");
  68. }
  69. auto c = _pool->borrow();
  70. pqxx::work txn { *c->c };
  71. pqxx::row r { txn.exec1("SELECT version FROM ztc_database") };
  72. int dbVersion = r[0].as<int>();
  73. txn.commit();
  74. if (dbVersion < DB_MINIMUM_VERSION) {
  75. fprintf(stderr, "Central database schema version too low. This controller version requires a minimum schema version of %d. Please upgrade your Central instance", DB_MINIMUM_VERSION);
  76. exit(1);
  77. }
  78. _pool->unborrow(c);
  79. if (_rc != NULL) {
  80. auto innerspan = tracer->StartSpan("cv1::CV1::configureRedis");
  81. auto innerscope = tracer->WithActiveSpan(innerspan);
  82. sw::redis::ConnectionOptions opts;
  83. sw::redis::ConnectionPoolOptions poolOpts;
  84. opts.host = _rc->hostname;
  85. opts.port = _rc->port;
  86. opts.password = _rc->password;
  87. opts.db = 0;
  88. opts.keep_alive = true;
  89. opts.connect_timeout = std::chrono::seconds(3);
  90. poolOpts.size = 25;
  91. poolOpts.wait_timeout = std::chrono::seconds(5);
  92. poolOpts.connection_lifetime = std::chrono::minutes(3);
  93. poolOpts.connection_idle_time = std::chrono::minutes(1);
  94. if (_rc->clusterMode) {
  95. innerspan->SetAttribute("cluster_mode", "true");
  96. fprintf(stderr, "Using Redis in Cluster Mode\n");
  97. _cluster = std::make_shared<sw::redis::RedisCluster>(opts, poolOpts);
  98. }
  99. else {
  100. innerspan->SetAttribute("cluster_mode", "false");
  101. fprintf(stderr, "Using Redis in Standalone Mode\n");
  102. _redis = std::make_shared<sw::redis::Redis>(opts, poolOpts);
  103. }
  104. }
  105. _readyLock.lock();
  106. fprintf(stderr, "[%s] NOTICE: %.10llx controller PostgreSQL waiting for initial data download..." ZT_EOL_S, ::_timestr(), (unsigned long long)_myAddress.toInt());
  107. _waitNoticePrinted = true;
  108. initializeNetworks();
  109. initializeMembers();
  110. _heartbeatThread = std::thread(&CV1::heartbeat, this);
  111. _membersDbWatcher = std::thread(&CV1::membersDbWatcher, this);
  112. _networksDbWatcher = std::thread(&CV1::networksDbWatcher, this);
  113. for (int i = 0; i < ZT_CENTRAL_CONTROLLER_COMMIT_THREADS; ++i) {
  114. _commitThread[i] = std::thread(&CV1::commitThread, this);
  115. }
  116. _onlineNotificationThread = std::thread(&CV1::onlineNotificationThread, this);
  117. configureSmee();
  118. }
  119. CV1::~CV1()
  120. {
  121. if (_smee != NULL) {
  122. smeeclient::smee_client_delete(_smee);
  123. _smee = NULL;
  124. }
  125. _run = 0;
  126. std::this_thread::sleep_for(std::chrono::milliseconds(100));
  127. _heartbeatThread.join();
  128. _membersDbWatcher.join();
  129. _networksDbWatcher.join();
  130. _commitQueue.stop();
  131. for (int i = 0; i < ZT_CENTRAL_CONTROLLER_COMMIT_THREADS; ++i) {
  132. _commitThread[i].join();
  133. }
  134. _onlineNotificationThread.join();
  135. }
  136. void CV1::configureSmee()
  137. {
  138. auto provider = opentelemetry::trace::Provider::GetTracerProvider();
  139. auto tracer = provider->GetTracer("cv1");
  140. auto span = tracer->StartSpan("cv1::configureSmee");
  141. auto scope = tracer->WithActiveSpan(span);
  142. const char* TEMPORAL_SCHEME = "ZT_TEMPORAL_SCHEME";
  143. const char* TEMPORAL_HOST = "ZT_TEMPORAL_HOST";
  144. const char* TEMPORAL_PORT = "ZT_TEMPORAL_PORT";
  145. const char* TEMPORAL_NAMESPACE = "ZT_TEMPORAL_NAMESPACE";
  146. const char* SMEE_TASK_QUEUE = "ZT_SMEE_TASK_QUEUE";
  147. const char* scheme = getenv(TEMPORAL_SCHEME);
  148. if (scheme == NULL) {
  149. scheme = "http";
  150. }
  151. const char* host = getenv(TEMPORAL_HOST);
  152. const char* port = getenv(TEMPORAL_PORT);
  153. const char* ns = getenv(TEMPORAL_NAMESPACE);
  154. const char* task_queue = getenv(SMEE_TASK_QUEUE);
  155. if (scheme != NULL && host != NULL && port != NULL && ns != NULL && task_queue != NULL) {
  156. fprintf(stderr, "creating smee client\n");
  157. std::string hostPort = std::string(scheme) + std::string("://") + std::string(host) + std::string(":") + std::string(port);
  158. this->_smee = smeeclient::smee_client_new(hostPort.c_str(), ns, task_queue);
  159. }
  160. else {
  161. fprintf(stderr, "Smee client not configured\n");
  162. }
  163. }
  164. bool CV1::waitForReady()
  165. {
  166. while (_ready < 2) {
  167. _readyLock.lock();
  168. _readyLock.unlock();
  169. }
  170. return true;
  171. }
  172. bool CV1::isReady()
  173. {
  174. return ((_ready == 2) && (_connected));
  175. }
  176. bool CV1::save(nlohmann::json& record, bool notifyListeners)
  177. {
  178. auto provider = opentelemetry::trace::Provider::GetTracerProvider();
  179. auto tracer = provider->GetTracer("cv1");
  180. auto span = tracer->StartSpan("cv1::save");
  181. auto scope = tracer->WithActiveSpan(span);
  182. bool modified = false;
  183. try {
  184. if (! record.is_object()) {
  185. fprintf(stderr, "record is not an object?!?\n");
  186. return false;
  187. }
  188. const std::string objtype = record["objtype"];
  189. if (objtype == "network") {
  190. // fprintf(stderr, "network save\n");
  191. const uint64_t nwid = OSUtils::jsonIntHex(record["id"], 0ULL);
  192. if (nwid) {
  193. nlohmann::json old;
  194. get(nwid, old);
  195. if ((! old.is_object()) || (! _compareRecords(old, record))) {
  196. record["revision"] = OSUtils::jsonInt(record["revision"], 0ULL) + 1ULL;
  197. _commitQueue.post(std::pair<nlohmann::json, bool>(record, notifyListeners));
  198. modified = true;
  199. }
  200. }
  201. }
  202. else if (objtype == "member") {
  203. std::string networkId = record["nwid"];
  204. std::string memberId = record["id"];
  205. const uint64_t nwid = OSUtils::jsonIntHex(record["nwid"], 0ULL);
  206. const uint64_t id = OSUtils::jsonIntHex(record["id"], 0ULL);
  207. // fprintf(stderr, "member save %s-%s\n", networkId.c_str(), memberId.c_str());
  208. if ((id) && (nwid)) {
  209. nlohmann::json network, old;
  210. get(nwid, network, id, old);
  211. if ((! old.is_object()) || (! _compareRecords(old, record))) {
  212. // fprintf(stderr, "commit queue post\n");
  213. record["revision"] = OSUtils::jsonInt(record["revision"], 0ULL) + 1ULL;
  214. _commitQueue.post(std::pair<nlohmann::json, bool>(record, notifyListeners));
  215. modified = true;
  216. }
  217. else {
  218. // fprintf(stderr, "no change\n");
  219. }
  220. }
  221. }
  222. else {
  223. fprintf(stderr, "uhh waaat\n");
  224. }
  225. }
  226. catch (std::exception& e) {
  227. fprintf(stderr, "Error on PostgreSQL::save: %s\n", e.what());
  228. }
  229. catch (...) {
  230. fprintf(stderr, "Unknown error on PostgreSQL::save\n");
  231. }
  232. return modified;
  233. }
  234. void CV1::eraseNetwork(const uint64_t networkId)
  235. {
  236. auto provider = opentelemetry::trace::Provider::GetTracerProvider();
  237. auto tracer = provider->GetTracer("cv1");
  238. auto span = tracer->StartSpan("cv1::eraseNetwork");
  239. auto scope = tracer->WithActiveSpan(span);
  240. char networkIdStr[17];
  241. span->SetAttribute("network_id", Utils::hex(networkId, networkIdStr));
  242. fprintf(stderr, "PostgreSQL::eraseNetwork\n");
  243. char tmp2[24];
  244. waitForReady();
  245. Utils::hex(networkId, tmp2);
  246. std::pair<nlohmann::json, bool> tmp;
  247. tmp.first["id"] = tmp2;
  248. tmp.first["objtype"] = "_delete_network";
  249. tmp.second = true;
  250. _commitQueue.post(tmp);
  251. nlohmann::json nullJson;
  252. _networkChanged(tmp.first, nullJson, true);
  253. }
  254. void CV1::eraseMember(const uint64_t networkId, const uint64_t memberId)
  255. {
  256. auto provider = opentelemetry::trace::Provider::GetTracerProvider();
  257. auto tracer = provider->GetTracer("cv1");
  258. auto span = tracer->StartSpan("cv1::eraseMember");
  259. auto scope = tracer->WithActiveSpan(span);
  260. char networkIdStr[17];
  261. char memberIdStr[11];
  262. span->SetAttribute("network_id", Utils::hex(networkId, networkIdStr));
  263. span->SetAttribute("member_id", Utils::hex10(memberId, memberIdStr));
  264. fprintf(stderr, "PostgreSQL::eraseMember\n");
  265. char tmp2[24];
  266. waitForReady();
  267. std::pair<nlohmann::json, bool> tmp, nw;
  268. Utils::hex(networkId, tmp2);
  269. tmp.first["nwid"] = tmp2;
  270. Utils::hex(memberId, tmp2);
  271. tmp.first["id"] = tmp2;
  272. tmp.first["objtype"] = "_delete_member";
  273. tmp.second = true;
  274. _commitQueue.post(tmp);
  275. nlohmann::json nullJson;
  276. _memberChanged(tmp.first, nullJson, true);
  277. }
  278. void CV1::nodeIsOnline(const uint64_t networkId, const uint64_t memberId, const InetAddress& physicalAddress, const char* osArch)
  279. {
  280. auto provider = opentelemetry::trace::Provider::GetTracerProvider();
  281. auto tracer = provider->GetTracer("cv1");
  282. auto span = tracer->StartSpan("cv1::nodeIsOnline");
  283. auto scope = tracer->WithActiveSpan(span);
  284. char networkIdStr[17];
  285. char memberIdStr[11];
  286. char ipStr[INET6_ADDRSTRLEN];
  287. span->SetAttribute("network_id", Utils::hex(networkId, networkIdStr));
  288. span->SetAttribute("member_id", Utils::hex10(memberId, memberIdStr));
  289. span->SetAttribute("physical_address", physicalAddress.toString(ipStr));
  290. span->SetAttribute("os_arch", osArch);
  291. std::lock_guard<std::mutex> l(_lastOnline_l);
  292. NodeOnlineRecord& i = _lastOnline[std::pair<uint64_t, uint64_t>(networkId, memberId)];
  293. i.lastSeen = OSUtils::now();
  294. if (physicalAddress) {
  295. i.physicalAddress = physicalAddress;
  296. }
  297. i.osArch = std::string(osArch);
  298. }
  299. void CV1::nodeIsOnline(const uint64_t networkId, const uint64_t memberId, const InetAddress& physicalAddress)
  300. {
  301. this->nodeIsOnline(networkId, memberId, physicalAddress, "unknown/unknown");
  302. }
  303. AuthInfo CV1::getSSOAuthInfo(const nlohmann::json& member, const std::string& redirectURL)
  304. {
  305. auto provider = opentelemetry::trace::Provider::GetTracerProvider();
  306. auto tracer = provider->GetTracer("cv1");
  307. auto span = tracer->StartSpan("cv1::getSSOAuthInfo");
  308. auto scope = tracer->WithActiveSpan(span);
  309. Metrics::db_get_sso_info++;
  310. // NONCE is just a random character string. no semantic meaning
  311. // state = HMAC SHA384 of Nonce based on shared sso key
  312. //
  313. // need nonce timeout in database? make sure it's used within X time
  314. // X is 5 minutes for now. Make configurable later?
  315. //
  316. // how do we tell when a nonce is used? if auth_expiration_time is set
  317. std::string networkId = member["nwid"];
  318. std::string memberId = member["id"];
  319. char authenticationURL[4096] = { 0 };
  320. AuthInfo info;
  321. info.enabled = true;
  322. // if (memberId == "a10dccea52" && networkId == "8056c2e21c24673d") {
  323. // fprintf(stderr, "invalid authinfo for grant's machine\n");
  324. // info.version=1;
  325. // return info;
  326. // }
  327. // fprintf(stderr, "PostgreSQL::updateMemberOnLoad: %s-%s\n", networkId.c_str(), memberId.c_str());
  328. std::shared_ptr<PostgresConnection> c;
  329. try {
  330. c = _pool->borrow();
  331. pqxx::work w(*c->c);
  332. char nonceBytes[16] = { 0 };
  333. std::string nonce = "";
  334. // check if the member exists first.
  335. pqxx::row count = w.exec_params1("SELECT count(id) FROM ztc_member WHERE id = $1 AND network_id = $2 AND deleted = false", memberId, networkId);
  336. if (count[0].as<int>() == 1) {
  337. // get active nonce, if exists.
  338. pqxx::result r = w.exec_params(
  339. "SELECT nonce FROM ztc_sso_expiry "
  340. "WHERE network_id = $1 AND member_id = $2 "
  341. "AND ((NOW() AT TIME ZONE 'UTC') <= authentication_expiry_time) AND ((NOW() AT TIME ZONE 'UTC') <= nonce_expiration)",
  342. networkId,
  343. memberId);
  344. if (r.size() == 0) {
  345. // no active nonce.
  346. // find an unused nonce, if one exists.
  347. pqxx::result r = w.exec_params(
  348. "SELECT nonce FROM ztc_sso_expiry "
  349. "WHERE network_id = $1 AND member_id = $2 "
  350. "AND authentication_expiry_time IS NULL AND ((NOW() AT TIME ZONE 'UTC') <= nonce_expiration)",
  351. networkId,
  352. memberId);
  353. if (r.size() == 1) {
  354. // we have an existing nonce. Use it
  355. nonce = r.at(0)[0].as<std::string>();
  356. Utils::unhex(nonce.c_str(), nonceBytes, sizeof(nonceBytes));
  357. }
  358. else if (r.empty()) {
  359. // create a nonce
  360. Utils::getSecureRandom(nonceBytes, 16);
  361. char nonceBuf[64] = { 0 };
  362. Utils::hex(nonceBytes, sizeof(nonceBytes), nonceBuf);
  363. nonce = std::string(nonceBuf);
  364. pqxx::result ir = w.exec_params0(
  365. "INSERT INTO ztc_sso_expiry "
  366. "(nonce, nonce_expiration, network_id, member_id) VALUES "
  367. "($1, TO_TIMESTAMP($2::double precision/1000), $3, $4)",
  368. nonce,
  369. OSUtils::now() + 300000,
  370. networkId,
  371. memberId);
  372. w.commit();
  373. }
  374. else {
  375. // > 1 ?!? Thats an error!
  376. fprintf(stderr, "> 1 unused nonce!\n");
  377. exit(6);
  378. }
  379. }
  380. else if (r.size() == 1) {
  381. nonce = r.at(0)[0].as<std::string>();
  382. Utils::unhex(nonce.c_str(), nonceBytes, sizeof(nonceBytes));
  383. }
  384. else {
  385. // more than 1 nonce in use? Uhhh...
  386. fprintf(stderr, "> 1 nonce in use for network member?!?\n");
  387. exit(7);
  388. }
  389. r = w.exec_params(
  390. "SELECT oc.client_id, oc.authorization_endpoint, oc.issuer, oc.provider, oc.sso_impl_version "
  391. "FROM ztc_network AS n "
  392. "INNER JOIN ztc_org o "
  393. " ON o.owner_id = n.owner_id "
  394. "LEFT OUTER JOIN ztc_network_oidc_config noc "
  395. " ON noc.network_id = n.id "
  396. "LEFT OUTER JOIN ztc_oidc_config oc "
  397. " ON noc.client_id = oc.client_id AND oc.org_id = o.org_id "
  398. "WHERE n.id = $1 AND n.sso_enabled = true",
  399. networkId);
  400. std::string client_id = "";
  401. std::string authorization_endpoint = "";
  402. std::string issuer = "";
  403. std::string provider = "";
  404. uint64_t sso_version = 0;
  405. if (r.size() == 1) {
  406. client_id = r.at(0)[0].as<std::optional<std::string> >().value_or("");
  407. authorization_endpoint = r.at(0)[1].as<std::optional<std::string> >().value_or("");
  408. issuer = r.at(0)[2].as<std::optional<std::string> >().value_or("");
  409. provider = r.at(0)[3].as<std::optional<std::string> >().value_or("");
  410. sso_version = r.at(0)[4].as<std::optional<uint64_t> >().value_or(1);
  411. }
  412. else if (r.size() > 1) {
  413. fprintf(stderr, "ERROR: More than one auth endpoint for an organization?!?!? NetworkID: %s\n", networkId.c_str());
  414. }
  415. else {
  416. fprintf(stderr, "No client or auth endpoint?!?\n");
  417. }
  418. info.version = sso_version;
  419. // no catch all else because we don't actually care if no records exist here. just continue as normal.
  420. if ((! client_id.empty()) && (! authorization_endpoint.empty())) {
  421. uint8_t state[48];
  422. HMACSHA384(_ssoPsk, nonceBytes, sizeof(nonceBytes), state);
  423. char state_hex[256];
  424. Utils::hex(state, 48, state_hex);
  425. if (info.version == 0) {
  426. char url[2048] = { 0 };
  427. OSUtils::ztsnprintf(
  428. url,
  429. sizeof(authenticationURL),
  430. "%s?response_type=id_token&response_mode=form_post&scope=openid+email+profile&redirect_uri=%s&nonce=%s&state=%s&client_id=%s",
  431. authorization_endpoint.c_str(),
  432. url_encode(redirectURL).c_str(),
  433. nonce.c_str(),
  434. state_hex,
  435. client_id.c_str());
  436. info.authenticationURL = std::string(url);
  437. }
  438. else if (info.version == 1) {
  439. info.ssoClientID = client_id;
  440. info.issuerURL = issuer;
  441. info.ssoProvider = provider;
  442. info.ssoNonce = nonce;
  443. info.ssoState = std::string(state_hex) + "_" + networkId;
  444. info.centralAuthURL = redirectURL;
  445. #ifdef ZT_DEBUG
  446. fprintf(
  447. stderr,
  448. "ssoClientID: %s\nissuerURL: %s\nssoNonce: %s\nssoState: %s\ncentralAuthURL: %s\nprovider: %s\n",
  449. info.ssoClientID.c_str(),
  450. info.issuerURL.c_str(),
  451. info.ssoNonce.c_str(),
  452. info.ssoState.c_str(),
  453. info.centralAuthURL.c_str(),
  454. provider.c_str());
  455. #endif
  456. }
  457. }
  458. else {
  459. fprintf(stderr, "client_id: %s\nauthorization_endpoint: %s\n", client_id.c_str(), authorization_endpoint.c_str());
  460. }
  461. }
  462. _pool->unborrow(c);
  463. }
  464. catch (std::exception& e) {
  465. span->SetStatus(opentelemetry::trace::StatusCode::kError, e.what());
  466. fprintf(stderr, "ERROR: Error updating member on load for network %s: %s\n", networkId.c_str(), e.what());
  467. }
  468. return info; // std::string(authenticationURL);
  469. }
  470. void CV1::initializeNetworks()
  471. {
  472. auto provider = opentelemetry::trace::Provider::GetTracerProvider();
  473. auto tracer = provider->GetTracer("cv1");
  474. auto span = tracer->StartSpan("cv1::initializeNetworks");
  475. auto scope = tracer->WithActiveSpan(span);
  476. try {
  477. std::string setKey = "networks:{" + _myAddressStr + "}";
  478. fprintf(stderr, "Initializing Networks...\n");
  479. if (_redisMemberStatus) {
  480. fprintf(stderr, "Init Redis for networks...\n");
  481. try {
  482. if (_rc->clusterMode) {
  483. _cluster->del(setKey);
  484. }
  485. else {
  486. _redis->del(setKey);
  487. }
  488. }
  489. catch (sw::redis::Error& e) {
  490. // ignore. if this key doesn't exist, there's no reason to delete it
  491. }
  492. }
  493. std::unordered_set<std::string> networkSet;
  494. char qbuf[2048] = { 0 };
  495. sprintf(
  496. qbuf,
  497. "SELECT n.id, (EXTRACT(EPOCH FROM n.creation_time AT TIME ZONE 'UTC')*1000)::bigint as creation_time, n.capabilities, "
  498. "n.enable_broadcast, (EXTRACT(EPOCH FROM n.last_modified AT TIME ZONE 'UTC')*1000)::bigint AS last_modified, n.mtu, n.multicast_limit, n.name, n.private, n.remote_trace_level, "
  499. "n.remote_trace_target, n.revision, n.rules, n.tags, n.v4_assign_mode, n.v6_assign_mode, n.sso_enabled, (CASE WHEN n.sso_enabled THEN noc.client_id ELSE NULL END) as client_id, "
  500. "(CASE WHEN n.sso_enabled THEN oc.authorization_endpoint ELSE NULL END) as authorization_endpoint, "
  501. "(CASE WHEN n.sso_enabled THEN oc.provider ELSE NULL END) as provider, d.domain, d.servers, "
  502. "ARRAY(SELECT CONCAT(host(ip_range_start),'|', host(ip_range_end)) FROM ztc_network_assignment_pool WHERE network_id = n.id) AS assignment_pool, "
  503. "ARRAY(SELECT CONCAT(host(address),'/',bits::text,'|',COALESCE(host(via), 'NULL'))FROM ztc_network_route WHERE network_id = n.id) AS routes "
  504. "FROM ztc_network n "
  505. "LEFT OUTER JOIN ztc_org o "
  506. " ON o.owner_id = n.owner_id "
  507. "LEFT OUTER JOIN ztc_network_oidc_config noc "
  508. " ON noc.network_id = n.id "
  509. "LEFT OUTER JOIN ztc_oidc_config oc "
  510. " ON noc.client_id = oc.client_id AND oc.org_id = o.org_id "
  511. "LEFT OUTER JOIN ztc_network_dns d "
  512. " ON d.network_id = n.id "
  513. "WHERE deleted = false AND controller_id = '%s'",
  514. _myAddressStr.c_str());
  515. auto c = _pool->borrow();
  516. auto c2 = _pool->borrow();
  517. pqxx::work w { *c->c };
  518. fprintf(stderr, "Load networks from psql...\n");
  519. auto stream = pqxx::stream_from::query(w, qbuf);
  520. std::tuple<
  521. std::string // network ID
  522. ,
  523. std::optional<int64_t> // creationTime
  524. ,
  525. std::optional<std::string> // capabilities
  526. ,
  527. std::optional<bool> // enableBroadcast
  528. ,
  529. std::optional<uint64_t> // lastModified
  530. ,
  531. std::optional<int> // mtu
  532. ,
  533. std::optional<int> // multicastLimit
  534. ,
  535. std::optional<std::string> // name
  536. ,
  537. bool // private
  538. ,
  539. std::optional<int> // remoteTraceLevel
  540. ,
  541. std::optional<std::string> // remoteTraceTarget
  542. ,
  543. std::optional<uint64_t> // revision
  544. ,
  545. std::optional<std::string> // rules
  546. ,
  547. std::optional<std::string> // tags
  548. ,
  549. std::optional<std::string> // v4AssignMode
  550. ,
  551. std::optional<std::string> // v6AssignMode
  552. ,
  553. std::optional<bool> // ssoEnabled
  554. ,
  555. std::optional<std::string> // clientId
  556. ,
  557. std::optional<std::string> // authorizationEndpoint
  558. ,
  559. std::optional<std::string> // ssoProvider
  560. ,
  561. std::optional<std::string> // domain
  562. ,
  563. std::optional<std::string> // servers
  564. ,
  565. std::string // assignmentPoolString
  566. ,
  567. std::string // routeString
  568. >
  569. row;
  570. uint64_t count = 0;
  571. auto tmp = std::chrono::high_resolution_clock::now();
  572. uint64_t total = 0;
  573. while (stream >> row) {
  574. auto start = std::chrono::high_resolution_clock::now();
  575. json empty;
  576. json config;
  577. initNetwork(config);
  578. std::string nwid = std::get<0>(row);
  579. std::optional<int64_t> creationTime = std::get<1>(row);
  580. std::optional<std::string> capabilities = std::get<2>(row);
  581. std::optional<bool> enableBroadcast = std::get<3>(row);
  582. std::optional<uint64_t> lastModified = std::get<4>(row);
  583. std::optional<int> mtu = std::get<5>(row);
  584. std::optional<int> multicastLimit = std::get<6>(row);
  585. std::optional<std::string> name = std::get<7>(row);
  586. bool isPrivate = std::get<8>(row);
  587. std::optional<int> remoteTraceLevel = std::get<9>(row);
  588. std::optional<std::string> remoteTraceTarget = std::get<10>(row);
  589. std::optional<uint64_t> revision = std::get<11>(row);
  590. std::optional<std::string> rules = std::get<12>(row);
  591. std::optional<std::string> tags = std::get<13>(row);
  592. std::optional<std::string> v4AssignMode = std::get<14>(row);
  593. std::optional<std::string> v6AssignMode = std::get<15>(row);
  594. std::optional<bool> ssoEnabled = std::get<16>(row);
  595. std::optional<std::string> clientId = std::get<17>(row);
  596. std::optional<std::string> authorizationEndpoint = std::get<18>(row);
  597. std::optional<std::string> ssoProvider = std::get<19>(row);
  598. std::optional<std::string> dnsDomain = std::get<20>(row);
  599. std::optional<std::string> dnsServers = std::get<21>(row);
  600. std::string assignmentPoolString = std::get<22>(row);
  601. std::string routesString = std::get<23>(row);
  602. config["id"] = nwid;
  603. config["nwid"] = nwid;
  604. config["creationTime"] = creationTime.value_or(0);
  605. config["capabilities"] = json::parse(capabilities.value_or("[]"));
  606. config["enableBroadcast"] = enableBroadcast.value_or(false);
  607. config["lastModified"] = lastModified.value_or(0);
  608. config["mtu"] = mtu.value_or(2800);
  609. config["multicastLimit"] = multicastLimit.value_or(64);
  610. config["name"] = name.value_or("");
  611. config["private"] = isPrivate;
  612. config["remoteTraceLevel"] = remoteTraceLevel.value_or(0);
  613. config["remoteTraceTarget"] = remoteTraceTarget.value_or("");
  614. config["revision"] = revision.value_or(0);
  615. config["rules"] = json::parse(rules.value_or("[]"));
  616. config["tags"] = json::parse(tags.value_or("[]"));
  617. config["v4AssignMode"] = json::parse(v4AssignMode.value_or("{}"));
  618. config["v6AssignMode"] = json::parse(v6AssignMode.value_or("{}"));
  619. config["ssoEnabled"] = ssoEnabled.value_or(false);
  620. config["objtype"] = "network";
  621. config["ipAssignmentPools"] = json::array();
  622. config["routes"] = json::array();
  623. config["clientId"] = clientId.value_or("");
  624. config["authorizationEndpoint"] = authorizationEndpoint.value_or("");
  625. config["provider"] = ssoProvider.value_or("");
  626. networkSet.insert(nwid);
  627. if (dnsDomain.has_value()) {
  628. std::string serverList = dnsServers.value();
  629. json obj;
  630. auto servers = json::array();
  631. if (serverList.rfind("{", 0) != std::string::npos) {
  632. serverList = serverList.substr(1, serverList.size() - 2);
  633. std::stringstream ss(serverList);
  634. while (ss.good()) {
  635. std::string server;
  636. std::getline(ss, server, ',');
  637. servers.push_back(server);
  638. }
  639. }
  640. obj["domain"] = dnsDomain.value();
  641. obj["servers"] = servers;
  642. config["dns"] = obj;
  643. }
  644. config["ipAssignmentPools"] = json::array();
  645. if (assignmentPoolString != "{}") {
  646. std::string tmp = assignmentPoolString.substr(1, assignmentPoolString.size() - 2);
  647. std::vector<std::string> assignmentPools = split(tmp, ',');
  648. for (auto it = assignmentPools.begin(); it != assignmentPools.end(); ++it) {
  649. std::vector<std::string> r = split(*it, '|');
  650. json ip;
  651. ip["ipRangeStart"] = r[0];
  652. ip["ipRangeEnd"] = r[1];
  653. config["ipAssignmentPools"].push_back(ip);
  654. }
  655. }
  656. config["routes"] = json::array();
  657. if (routesString != "{}") {
  658. std::string tmp = routesString.substr(1, routesString.size() - 2);
  659. std::vector<std::string> routes = split(tmp, ',');
  660. for (auto it = routes.begin(); it != routes.end(); ++it) {
  661. std::vector<std::string> r = split(*it, '|');
  662. json route;
  663. route["target"] = r[0];
  664. route["via"] = ((route["via"] == "NULL") ? nullptr : r[1]);
  665. config["routes"].push_back(route);
  666. }
  667. }
  668. Metrics::network_count++;
  669. _networkChanged(empty, config, false);
  670. auto end = std::chrono::high_resolution_clock::now();
  671. auto dur = std::chrono::duration_cast<std::chrono::microseconds>(end - start);
  672. ;
  673. total += dur.count();
  674. ++count;
  675. if (count > 0 && count % 10000 == 0) {
  676. fprintf(stderr, "Averaging %llu us per network\n", (total / count));
  677. }
  678. }
  679. if (count > 0) {
  680. fprintf(stderr, "Took %llu us per network to load\n", (total / count));
  681. }
  682. stream.complete();
  683. w.commit();
  684. _pool->unborrow(c2);
  685. _pool->unborrow(c);
  686. fprintf(stderr, "done.\n");
  687. if (! networkSet.empty()) {
  688. if (_redisMemberStatus) {
  689. fprintf(stderr, "adding networks to redis...\n");
  690. if (_rc->clusterMode) {
  691. auto tx = _cluster->transaction(_myAddressStr, true, false);
  692. uint64_t count = 0;
  693. for (std::string nwid : networkSet) {
  694. tx.sadd(setKey, nwid);
  695. if (++count % 30000 == 0) {
  696. tx.exec();
  697. tx = _cluster->transaction(_myAddressStr, true, false);
  698. }
  699. }
  700. tx.exec();
  701. }
  702. else {
  703. auto tx = _redis->transaction(true, false);
  704. uint64_t count = 0;
  705. for (std::string nwid : networkSet) {
  706. tx.sadd(setKey, nwid);
  707. if (++count % 30000 == 0) {
  708. tx.exec();
  709. tx = _redis->transaction(true, false);
  710. }
  711. }
  712. tx.exec();
  713. }
  714. fprintf(stderr, "done.\n");
  715. }
  716. }
  717. if (++this->_ready == 2) {
  718. if (_waitNoticePrinted) {
  719. fprintf(stderr, "[%s] NOTICE: %.10llx controller PostgreSQL data download complete." ZT_EOL_S, _timestr(), (unsigned long long)_myAddress.toInt());
  720. }
  721. _readyLock.unlock();
  722. }
  723. fprintf(stderr, "network init done.\n");
  724. }
  725. catch (sw::redis::Error& e) {
  726. span->SetStatus(opentelemetry::trace::StatusCode::kError, e.what());
  727. fprintf(stderr, "ERROR: Error initializing networks in Redis: %s\n", e.what());
  728. std::this_thread::sleep_for(std::chrono::milliseconds(5000));
  729. exit(-1);
  730. }
  731. catch (std::exception& e) {
  732. span->SetStatus(opentelemetry::trace::StatusCode::kError, e.what());
  733. fprintf(stderr, "ERROR: Error initializing networks: %s\n", e.what());
  734. std::this_thread::sleep_for(std::chrono::milliseconds(5000));
  735. exit(-1);
  736. }
  737. }
  738. void CV1::initializeMembers()
  739. {
  740. auto provider = opentelemetry::trace::Provider::GetTracerProvider();
  741. auto tracer = provider->GetTracer("cv1");
  742. auto span = tracer->StartSpan("cv1::initializeMembers");
  743. auto scope = tracer->WithActiveSpan(span);
  744. std::string memberId;
  745. std::string networkId;
  746. try {
  747. std::unordered_map<std::string, std::string> networkMembers;
  748. fprintf(stderr, "Initializing Members...\n");
  749. std::string setKeyBase = "network-nodes-all:{" + _myAddressStr + "}:";
  750. if (_redisMemberStatus) {
  751. fprintf(stderr, "Initialize Redis for members...\n");
  752. std::unique_lock<std::shared_mutex> l(_networks_l);
  753. std::unordered_set<std::string> deletes;
  754. for (auto it : _networks) {
  755. uint64_t nwid_i = it.first;
  756. char nwidTmp[64] = { 0 };
  757. OSUtils::ztsnprintf(nwidTmp, sizeof(nwidTmp), "%.16llx", nwid_i);
  758. std::string nwid(nwidTmp);
  759. std::string key = setKeyBase + nwid;
  760. deletes.insert(key);
  761. }
  762. if (! deletes.empty()) {
  763. try {
  764. if (_rc->clusterMode) {
  765. auto tx = _cluster->transaction(_myAddressStr, true, false);
  766. for (std::string k : deletes) {
  767. tx.del(k);
  768. }
  769. tx.exec();
  770. }
  771. else {
  772. auto tx = _redis->transaction(true, false);
  773. for (std::string k : deletes) {
  774. tx.del(k);
  775. }
  776. tx.exec();
  777. }
  778. }
  779. catch (sw::redis::Error& e) {
  780. // ignore
  781. }
  782. }
  783. }
  784. char qbuf[2048];
  785. sprintf(
  786. qbuf,
  787. "SELECT m.id, m.network_id, m.active_bridge, m.authorized, m.capabilities, "
  788. "(EXTRACT(EPOCH FROM m.creation_time AT TIME ZONE 'UTC')*1000)::bigint, m.identity, "
  789. "(EXTRACT(EPOCH FROM m.last_authorized_time AT TIME ZONE 'UTC')*1000)::bigint, "
  790. "(EXTRACT(EPOCH FROM m.last_deauthorized_time AT TIME ZONE 'UTC')*1000)::bigint, "
  791. "m.remote_trace_level, m.remote_trace_target, m.tags, m.v_major, m.v_minor, m.v_rev, m.v_proto, "
  792. "m.no_auto_assign_ips, m.revision, m.sso_exempt, "
  793. "(CASE WHEN n.sso_enabled = TRUE AND m.sso_exempt = FALSE THEN "
  794. " ( "
  795. " SELECT (EXTRACT(EPOCH FROM e.authentication_expiry_time)*1000)::bigint "
  796. " FROM ztc_sso_expiry e "
  797. " INNER JOIN ztc_network n1 "
  798. " ON n1.id = e.network_id AND n1.deleted = TRUE "
  799. " WHERE e.network_id = m.network_id AND e.member_id = m.id AND n.sso_enabled = TRUE AND e.authentication_expiry_time IS NOT NULL "
  800. " ORDER BY e.authentication_expiry_time DESC LIMIT 1 "
  801. " ) "
  802. " ELSE NULL "
  803. " END) AS authentication_expiry_time, "
  804. "ARRAY(SELECT DISTINCT address FROM ztc_member_ip_assignment WHERE member_id = m.id AND network_id = m.network_id) AS assigned_addresses "
  805. "FROM ztc_member m "
  806. "INNER JOIN ztc_network n "
  807. " ON n.id = m.network_id "
  808. "WHERE n.controller_id = '%s' AND n.deleted = FALSE AND m.deleted = FALSE",
  809. _myAddressStr.c_str());
  810. auto c = _pool->borrow();
  811. auto c2 = _pool->borrow();
  812. pqxx::work w { *c->c };
  813. fprintf(stderr, "Load members from psql...\n");
  814. auto stream = pqxx::stream_from::query(w, qbuf);
  815. std::tuple<
  816. std::string // memberId
  817. ,
  818. std::string // memberId
  819. ,
  820. std::optional<bool> // activeBridge
  821. ,
  822. std::optional<bool> // authorized
  823. ,
  824. std::optional<std::string> // capabilities
  825. ,
  826. std::optional<uint64_t> // creationTime
  827. ,
  828. std::optional<std::string> // identity
  829. ,
  830. std::optional<uint64_t> // lastAuthorizedTime
  831. ,
  832. std::optional<uint64_t> // lastDeauthorizedTime
  833. ,
  834. std::optional<int> // remoteTraceLevel
  835. ,
  836. std::optional<std::string> // remoteTraceTarget
  837. ,
  838. std::optional<std::string> // tags
  839. ,
  840. std::optional<int> // vMajor
  841. ,
  842. std::optional<int> // vMinor
  843. ,
  844. std::optional<int> // vRev
  845. ,
  846. std::optional<int> // vProto
  847. ,
  848. std::optional<bool> // noAutoAssignIps
  849. ,
  850. std::optional<uint64_t> // revision
  851. ,
  852. std::optional<bool> // ssoExempt
  853. ,
  854. std::optional<uint64_t> // authenticationExpiryTime
  855. ,
  856. std::string // assignedAddresses
  857. >
  858. row;
  859. uint64_t count = 0;
  860. auto tmp = std::chrono::high_resolution_clock::now();
  861. uint64_t total = 0;
  862. while (stream >> row) {
  863. auto start = std::chrono::high_resolution_clock::now();
  864. json empty;
  865. json config;
  866. initMember(config);
  867. memberId = std::get<0>(row);
  868. networkId = std::get<1>(row);
  869. std::optional<bool> activeBridge = std::get<2>(row);
  870. std::optional<bool> authorized = std::get<3>(row);
  871. std::optional<std::string> capabilities = std::get<4>(row);
  872. std::optional<uint64_t> creationTime = std::get<5>(row);
  873. std::optional<std::string> identity = std::get<6>(row);
  874. std::optional<uint64_t> lastAuthorizedTime = std::get<7>(row);
  875. std::optional<uint64_t> lastDeauthorizedTime = std::get<8>(row);
  876. std::optional<int> remoteTraceLevel = std::get<9>(row);
  877. std::optional<std::string> remoteTraceTarget = std::get<10>(row);
  878. std::optional<std::string> tags = std::get<11>(row);
  879. std::optional<int> vMajor = std::get<12>(row);
  880. std::optional<int> vMinor = std::get<13>(row);
  881. std::optional<int> vRev = std::get<14>(row);
  882. std::optional<int> vProto = std::get<15>(row);
  883. std::optional<bool> noAutoAssignIps = std::get<16>(row);
  884. std::optional<uint64_t> revision = std::get<17>(row);
  885. std::optional<bool> ssoExempt = std::get<18>(row);
  886. std::optional<uint64_t> authenticationExpiryTime = std::get<19>(row);
  887. std::string assignedAddresses = std::get<20>(row);
  888. networkMembers.insert(std::pair<std::string, std::string>(setKeyBase + networkId, memberId));
  889. config["id"] = memberId;
  890. config["address"] = memberId;
  891. config["nwid"] = networkId;
  892. config["activeBridge"] = activeBridge.value_or(false);
  893. config["authorized"] = authorized.value_or(false);
  894. config["capabilities"] = json::parse(capabilities.value_or("[]"));
  895. config["creationTime"] = creationTime.value_or(0);
  896. config["identity"] = identity.value_or("");
  897. config["lastAuthorizedTime"] = lastAuthorizedTime.value_or(0);
  898. config["lastDeauthorizedTime"] = lastDeauthorizedTime.value_or(0);
  899. config["remoteTraceLevel"] = remoteTraceLevel.value_or(0);
  900. config["remoteTraceTarget"] = remoteTraceTarget.value_or("");
  901. config["tags"] = json::parse(tags.value_or("[]"));
  902. config["vMajor"] = vMajor.value_or(-1);
  903. config["vMinor"] = vMinor.value_or(-1);
  904. config["vRev"] = vRev.value_or(-1);
  905. config["vProto"] = vProto.value_or(-1);
  906. config["noAutoAssignIps"] = noAutoAssignIps.value_or(false);
  907. config["revision"] = revision.value_or(0);
  908. config["ssoExempt"] = ssoExempt.value_or(false);
  909. config["authenticationExpiryTime"] = authenticationExpiryTime.value_or(0);
  910. config["objtype"] = "member";
  911. config["ipAssignments"] = json::array();
  912. if (assignedAddresses != "{}") {
  913. std::string tmp = assignedAddresses.substr(1, assignedAddresses.size() - 2);
  914. std::vector<std::string> addrs = split(tmp, ',');
  915. for (auto it = addrs.begin(); it != addrs.end(); ++it) {
  916. config["ipAssignments"].push_back(*it);
  917. }
  918. }
  919. Metrics::member_count++;
  920. _memberChanged(empty, config, false);
  921. memberId = "";
  922. networkId = "";
  923. auto end = std::chrono::high_resolution_clock::now();
  924. auto dur = std::chrono::duration_cast<std::chrono::microseconds>(end - start);
  925. total += dur.count();
  926. ++count;
  927. if (count > 0 && count % 10000 == 0) {
  928. fprintf(stderr, "Averaging %llu us per member\n", (total / count));
  929. }
  930. }
  931. if (count > 0) {
  932. fprintf(stderr, "Took %llu us per member to load\n", (total / count));
  933. }
  934. stream.complete();
  935. w.commit();
  936. _pool->unborrow(c2);
  937. _pool->unborrow(c);
  938. fprintf(stderr, "done.\n");
  939. if (! networkMembers.empty()) {
  940. if (_redisMemberStatus) {
  941. fprintf(stderr, "Load member data into redis...\n");
  942. if (_rc->clusterMode) {
  943. auto tx = _cluster->transaction(_myAddressStr, true, false);
  944. uint64_t count = 0;
  945. for (auto it : networkMembers) {
  946. tx.sadd(it.first, it.second);
  947. if (++count % 30000 == 0) {
  948. tx.exec();
  949. tx = _cluster->transaction(_myAddressStr, true, false);
  950. }
  951. }
  952. tx.exec();
  953. }
  954. else {
  955. auto tx = _redis->transaction(true, false);
  956. uint64_t count = 0;
  957. for (auto it : networkMembers) {
  958. tx.sadd(it.first, it.second);
  959. if (++count % 30000 == 0) {
  960. tx.exec();
  961. tx = _redis->transaction(true, false);
  962. }
  963. }
  964. tx.exec();
  965. }
  966. fprintf(stderr, "done.\n");
  967. }
  968. }
  969. fprintf(stderr, "Done loading members...\n");
  970. if (++this->_ready == 2) {
  971. if (_waitNoticePrinted) {
  972. fprintf(stderr, "[%s] NOTICE: %.10llx controller PostgreSQL data download complete." ZT_EOL_S, _timestr(), (unsigned long long)_myAddress.toInt());
  973. }
  974. _readyLock.unlock();
  975. }
  976. }
  977. catch (sw::redis::Error& e) {
  978. span->SetStatus(opentelemetry::trace::StatusCode::kError, e.what());
  979. fprintf(stderr, "ERROR: Error initializing members (redis): %s\n", e.what());
  980. exit(-1);
  981. }
  982. catch (std::exception& e) {
  983. span->SetStatus(opentelemetry::trace::StatusCode::kError, e.what());
  984. fprintf(stderr, "ERROR: Error initializing member: %s-%s %s\n", networkId.c_str(), memberId.c_str(), e.what());
  985. exit(-1);
  986. }
  987. }
  988. void CV1::heartbeat()
  989. {
  990. char publicId[1024];
  991. char hostnameTmp[1024];
  992. _myId.toString(false, publicId);
  993. if (gethostname(hostnameTmp, sizeof(hostnameTmp)) != 0) {
  994. hostnameTmp[0] = (char)0;
  995. }
  996. else {
  997. for (int i = 0; i < (int)sizeof(hostnameTmp); ++i) {
  998. if ((hostnameTmp[i] == '.') || (hostnameTmp[i] == 0)) {
  999. hostnameTmp[i] = (char)0;
  1000. break;
  1001. }
  1002. }
  1003. }
  1004. const char* controllerId = _myAddressStr.c_str();
  1005. const char* publicIdentity = publicId;
  1006. const char* hostname = hostnameTmp;
  1007. while (_run == 1) {
  1008. auto provider = opentelemetry::trace::Provider::GetTracerProvider();
  1009. auto tracer = provider->GetTracer("cv1");
  1010. auto span = tracer->StartSpan("cv1::heartbeat");
  1011. auto scope = tracer->WithActiveSpan(span);
  1012. // fprintf(stderr, "%s: heartbeat\n", controllerId);
  1013. auto c = _pool->borrow();
  1014. int64_t ts = OSUtils::now();
  1015. if (c->c) {
  1016. std::string major = std::to_string(ZEROTIER_ONE_VERSION_MAJOR);
  1017. std::string minor = std::to_string(ZEROTIER_ONE_VERSION_MINOR);
  1018. std::string rev = std::to_string(ZEROTIER_ONE_VERSION_REVISION);
  1019. std::string build = std::to_string(ZEROTIER_ONE_VERSION_BUILD);
  1020. std::string now = std::to_string(ts);
  1021. std::string host_port = std::to_string(_listenPort);
  1022. std::string use_redis = (_rc != NULL) ? "true" : "false";
  1023. std::string redis_mem_status = (_redisMemberStatus) ? "true" : "false";
  1024. try {
  1025. pqxx::work w { *c->c };
  1026. pqxx::result res = w.exec0(
  1027. "INSERT INTO ztc_controller (id, cluster_host, last_alive, public_identity, v_major, v_minor, v_rev, v_build, host_port, use_redis, redis_member_status) "
  1028. "VALUES ("
  1029. + w.quote(controllerId) + ", " + w.quote(hostname) + ", TO_TIMESTAMP(" + now + "::double precision/1000), " + w.quote(publicIdentity) + ", " + major + ", " + minor + ", " + rev + ", " + build + ", " + host_port + ", "
  1030. + use_redis + ", " + redis_mem_status
  1031. + ") "
  1032. "ON CONFLICT (id) DO UPDATE SET cluster_host = EXCLUDED.cluster_host, last_alive = EXCLUDED.last_alive, "
  1033. "public_identity = EXCLUDED.public_identity, v_major = EXCLUDED.v_major, v_minor = EXCLUDED.v_minor, "
  1034. "v_rev = EXCLUDED.v_rev, v_build = EXCLUDED.v_rev, host_port = EXCLUDED.host_port, "
  1035. "use_redis = EXCLUDED.use_redis, redis_member_status = EXCLUDED.redis_member_status");
  1036. w.commit();
  1037. }
  1038. catch (std::exception& e) {
  1039. fprintf(stderr, "%s: Heartbeat update failed: %s\n", controllerId, e.what());
  1040. span->End();
  1041. std::this_thread::sleep_for(std::chrono::milliseconds(1000));
  1042. continue;
  1043. }
  1044. }
  1045. _pool->unborrow(c);
  1046. try {
  1047. if (_redisMemberStatus) {
  1048. if (_rc->clusterMode) {
  1049. _cluster->zadd("controllers", "controllerId", ts);
  1050. }
  1051. else {
  1052. _redis->zadd("controllers", "controllerId", ts);
  1053. }
  1054. }
  1055. }
  1056. catch (sw::redis::Error& e) {
  1057. fprintf(stderr, "ERROR: Redis error in heartbeat thread: %s\n", e.what());
  1058. }
  1059. span->End();
  1060. std::this_thread::sleep_for(std::chrono::milliseconds(1000));
  1061. }
  1062. fprintf(stderr, "Exited heartbeat thread\n");
  1063. }
  1064. void CV1::membersDbWatcher()
  1065. {
  1066. if (_rc) {
  1067. _membersWatcher_Redis();
  1068. }
  1069. else {
  1070. _membersWatcher_Postgres();
  1071. }
  1072. if (_run == 1) {
  1073. fprintf(stderr, "ERROR: %s membersDbWatcher should still be running! Exiting Controller.\n", _myAddressStr.c_str());
  1074. exit(9);
  1075. }
  1076. fprintf(stderr, "Exited membersDbWatcher\n");
  1077. }
  1078. void CV1::_membersWatcher_Postgres()
  1079. {
  1080. auto c = _pool->borrow();
  1081. std::string stream = "member_" + _myAddressStr;
  1082. fprintf(stderr, "Listening to member stream: %s\n", stream.c_str());
  1083. MemberNotificationReceiver<CV1> m(this, *c->c, stream);
  1084. while (_run == 1) {
  1085. c->c->await_notification(5, 0);
  1086. }
  1087. _pool->unborrow(c);
  1088. }
  1089. void CV1::_membersWatcher_Redis()
  1090. {
  1091. char buf[11] = { 0 };
  1092. std::string key = "member-stream:{" + std::string(_myAddress.toString(buf)) + "}";
  1093. std::string lastID = "0";
  1094. fprintf(stderr, "Listening to member stream: %s\n", key.c_str());
  1095. while (_run == 1) {
  1096. auto provider = opentelemetry::trace::Provider::GetTracerProvider();
  1097. auto tracer = provider->GetTracer("cv1");
  1098. auto span = tracer->StartSpan("cv1::_membersWatcher_Redis");
  1099. auto scope = tracer->WithActiveSpan(span);
  1100. try {
  1101. json tmp;
  1102. std::unordered_map<std::string, ItemStream> result;
  1103. if (_rc->clusterMode) {
  1104. _cluster->xread(key, lastID, std::chrono::seconds(1), 0, std::inserter(result, result.end()));
  1105. }
  1106. else {
  1107. _redis->xread(key, lastID, std::chrono::seconds(1), 0, std::inserter(result, result.end()));
  1108. }
  1109. if (! result.empty()) {
  1110. for (auto element : result) {
  1111. #ifdef REDIS_TRACE
  1112. fprintf(stdout, "Received notification from: %s\n", element.first.c_str());
  1113. #endif
  1114. for (auto rec : element.second) {
  1115. std::string id = rec.first;
  1116. auto attrs = rec.second;
  1117. #ifdef REDIS_TRACE
  1118. fprintf(stdout, "Record ID: %s\n", id.c_str());
  1119. fprintf(stdout, "attrs len: %lu\n", attrs.size());
  1120. #endif
  1121. for (auto a : attrs) {
  1122. #ifdef REDIS_TRACE
  1123. fprintf(stdout, "key: %s\nvalue: %s\n", a.first.c_str(), a.second.c_str());
  1124. #endif
  1125. try {
  1126. tmp = json::parse(a.second);
  1127. json& ov = tmp["old_val"];
  1128. json& nv = tmp["new_val"];
  1129. json oldConfig, newConfig;
  1130. if (ov.is_object())
  1131. oldConfig = ov;
  1132. if (nv.is_object())
  1133. newConfig = nv;
  1134. if (oldConfig.is_object() || newConfig.is_object()) {
  1135. _memberChanged(oldConfig, newConfig, (this->_ready >= 2));
  1136. }
  1137. }
  1138. catch (...) {
  1139. fprintf(stderr, "json parse error in _membersWatcher_Redis: %s\n", a.second.c_str());
  1140. }
  1141. }
  1142. if (_rc->clusterMode) {
  1143. _cluster->xdel(key, id);
  1144. }
  1145. else {
  1146. _redis->xdel(key, id);
  1147. }
  1148. lastID = id;
  1149. Metrics::redis_mem_notification++;
  1150. }
  1151. }
  1152. }
  1153. }
  1154. catch (sw::redis::Error& e) {
  1155. span->SetStatus(opentelemetry::trace::StatusCode::kError, e.what());
  1156. fprintf(stderr, "Error in Redis members watcher: %s\n", e.what());
  1157. }
  1158. }
  1159. fprintf(stderr, "membersWatcher ended\n");
  1160. }
  1161. void CV1::networksDbWatcher()
  1162. {
  1163. if (_rc) {
  1164. _networksWatcher_Redis();
  1165. }
  1166. else {
  1167. _networksWatcher_Postgres();
  1168. }
  1169. if (_run == 1) {
  1170. fprintf(stderr, "ERROR: %s networksDbWatcher should still be running! Exiting Controller.\n", _myAddressStr.c_str());
  1171. exit(8);
  1172. }
  1173. fprintf(stderr, "Exited networksDbWatcher\n");
  1174. }
  1175. void CV1::_networksWatcher_Postgres()
  1176. {
  1177. std::string stream = "network_" + _myAddressStr;
  1178. fprintf(stderr, "Listening to member stream: %s\n", stream.c_str());
  1179. auto c = _pool->borrow();
  1180. NetworkNotificationReceiver<CV1> n(this, *c->c, stream);
  1181. while (_run == 1) {
  1182. auto provider = opentelemetry::trace::Provider::GetTracerProvider();
  1183. auto tracer = provider->GetTracer("cv1");
  1184. auto span = tracer->StartSpan("cv1::_networksWatcher_Postgres");
  1185. auto scope = tracer->WithActiveSpan(span);
  1186. c->c->await_notification(5, 0);
  1187. }
  1188. }
  1189. void CV1::_networksWatcher_Redis()
  1190. {
  1191. char buf[11] = { 0 };
  1192. std::string key = "network-stream:{" + std::string(_myAddress.toString(buf)) + "}";
  1193. std::string lastID = "0";
  1194. while (_run == 1) {
  1195. auto provider = opentelemetry::trace::Provider::GetTracerProvider();
  1196. auto tracer = provider->GetTracer("cv1");
  1197. auto span = tracer->StartSpan("cv1::_networksWatcher_Redis");
  1198. auto scope = tracer->WithActiveSpan(span);
  1199. try {
  1200. json tmp;
  1201. std::unordered_map<std::string, ItemStream> result;
  1202. if (_rc->clusterMode) {
  1203. _cluster->xread(key, lastID, std::chrono::seconds(1), 0, std::inserter(result, result.end()));
  1204. }
  1205. else {
  1206. _redis->xread(key, lastID, std::chrono::seconds(1), 0, std::inserter(result, result.end()));
  1207. }
  1208. if (! result.empty()) {
  1209. for (auto element : result) {
  1210. #ifdef REDIS_TRACE
  1211. fprintf(stdout, "Received notification from: %s\n", element.first.c_str());
  1212. #endif
  1213. for (auto rec : element.second) {
  1214. std::string id = rec.first;
  1215. auto attrs = rec.second;
  1216. #ifdef REDIS_TRACE
  1217. fprintf(stdout, "Record ID: %s\n", id.c_str());
  1218. fprintf(stdout, "attrs len: %lu\n", attrs.size());
  1219. #endif
  1220. for (auto a : attrs) {
  1221. #ifdef REDIS_TRACE
  1222. fprintf(stdout, "key: %s\nvalue: %s\n", a.first.c_str(), a.second.c_str());
  1223. #endif
  1224. try {
  1225. tmp = json::parse(a.second);
  1226. json& ov = tmp["old_val"];
  1227. json& nv = tmp["new_val"];
  1228. json oldConfig, newConfig;
  1229. if (ov.is_object())
  1230. oldConfig = ov;
  1231. if (nv.is_object())
  1232. newConfig = nv;
  1233. if (oldConfig.is_object() || newConfig.is_object()) {
  1234. _networkChanged(oldConfig, newConfig, (this->_ready >= 2));
  1235. }
  1236. }
  1237. catch (std::exception& e) {
  1238. fprintf(stderr, "json parse error in networkWatcher_Redis: what: %s json: %s\n", e.what(), a.second.c_str());
  1239. }
  1240. }
  1241. if (_rc->clusterMode) {
  1242. _cluster->xdel(key, id);
  1243. }
  1244. else {
  1245. _redis->xdel(key, id);
  1246. }
  1247. lastID = id;
  1248. }
  1249. Metrics::redis_net_notification++;
  1250. }
  1251. }
  1252. }
  1253. catch (sw::redis::Error& e) {
  1254. span->SetStatus(opentelemetry::trace::StatusCode::kError, e.what());
  1255. fprintf(stderr, "Error in Redis networks watcher: %s\n", e.what());
  1256. }
  1257. }
  1258. fprintf(stderr, "networksWatcher ended\n");
  1259. }
  1260. void CV1::commitThread()
  1261. {
  1262. fprintf(stderr, "%s: commitThread start\n", _myAddressStr.c_str());
  1263. std::pair<nlohmann::json, bool> qitem;
  1264. while (_commitQueue.get(qitem) && (_run == 1)) {
  1265. auto provider = opentelemetry::trace::Provider::GetTracerProvider();
  1266. auto tracer = provider->GetTracer("cv1");
  1267. auto span = tracer->StartSpan("cv1::commitThread");
  1268. auto scope = tracer->WithActiveSpan(span);
  1269. // fprintf(stderr, "commitThread tick\n");
  1270. if (! qitem.first.is_object()) {
  1271. fprintf(stderr, "not an object\n");
  1272. continue;
  1273. }
  1274. std::shared_ptr<PostgresConnection> c;
  1275. try {
  1276. c = _pool->borrow();
  1277. }
  1278. catch (std::exception& e) {
  1279. fprintf(stderr, "ERROR: %s\n", e.what());
  1280. continue;
  1281. }
  1282. if (! c) {
  1283. fprintf(stderr, "Error getting database connection\n");
  1284. continue;
  1285. }
  1286. Metrics::pgsql_commit_ticks++;
  1287. try {
  1288. nlohmann::json& config = (qitem.first);
  1289. const std::string objtype = config["objtype"];
  1290. if (objtype == "member") {
  1291. auto mspan = tracer->StartSpan("cv1::commitThread::member");
  1292. auto mscope = tracer->WithActiveSpan(mspan);
  1293. // fprintf(stderr, "%s: commitThread: member\n", _myAddressStr.c_str());
  1294. std::string memberId;
  1295. std::string networkId;
  1296. try {
  1297. pqxx::work w(*c->c);
  1298. memberId = config["id"];
  1299. networkId = config["nwid"];
  1300. std::string target = "NULL";
  1301. if (! config["remoteTraceTarget"].is_null()) {
  1302. target = config["remoteTraceTarget"];
  1303. }
  1304. pqxx::row nwrow = w.exec_params1("SELECT COUNT(id) FROM ztc_network WHERE id = $1", networkId);
  1305. int nwcount = nwrow[0].as<int>();
  1306. if (nwcount != 1) {
  1307. fprintf(stderr, "network %s does not exist. skipping member upsert\n", networkId.c_str());
  1308. w.abort();
  1309. _pool->unborrow(c);
  1310. continue;
  1311. }
  1312. pqxx::row mrow = w.exec_params1("SELECT COUNT(id) FROM ztc_member WHERE id = $1 AND network_id = $2", memberId, networkId);
  1313. int membercount = mrow[0].as<int>();
  1314. bool isNewMember = false;
  1315. if (membercount == 0) {
  1316. // new member
  1317. isNewMember = true;
  1318. pqxx::result res = w.exec_params0(
  1319. "INSERT INTO ztc_member (id, network_id, active_bridge, authorized, capabilities, "
  1320. "identity, last_authorized_time, last_deauthorized_time, no_auto_assign_ips, "
  1321. "remote_trace_level, remote_trace_target, revision, tags, v_major, v_minor, v_rev, v_proto) "
  1322. "VALUES ($1, $2, $3, $4, $5, $6, "
  1323. "TO_TIMESTAMP($7::double precision/1000), TO_TIMESTAMP($8::double precision/1000), "
  1324. "$9, $10, $11, $12, $13, $14, $15, $16, $17)",
  1325. memberId,
  1326. networkId,
  1327. (bool)config["activeBridge"],
  1328. (bool)config["authorized"],
  1329. OSUtils::jsonDump(config["capabilities"], -1),
  1330. OSUtils::jsonString(config["identity"], ""),
  1331. (uint64_t)config["lastAuthorizedTime"],
  1332. (uint64_t)config["lastDeauthorizedTime"],
  1333. (bool)config["noAutoAssignIps"],
  1334. (int)config["remoteTraceLevel"],
  1335. target,
  1336. (uint64_t)config["revision"],
  1337. OSUtils::jsonDump(config["tags"], -1),
  1338. (int)config["vMajor"],
  1339. (int)config["vMinor"],
  1340. (int)config["vRev"],
  1341. (int)config["vProto"]);
  1342. }
  1343. else {
  1344. // existing member
  1345. pqxx::result res = w.exec_params0(
  1346. "UPDATE ztc_member "
  1347. "SET active_bridge = $3, authorized = $4, capabilities = $5, identity = $6, "
  1348. "last_authorized_time = TO_TIMESTAMP($7::double precision/1000), "
  1349. "last_deauthorized_time = TO_TIMESTAMP($8::double precision/1000), "
  1350. "no_auto_assign_ips = $9, remote_trace_level = $10, remote_trace_target= $11, "
  1351. "revision = $12, tags = $13, v_major = $14, v_minor = $15, v_rev = $16, v_proto = $17 "
  1352. "WHERE id = $1 AND network_id = $2",
  1353. memberId,
  1354. networkId,
  1355. (bool)config["activeBridge"],
  1356. (bool)config["authorized"],
  1357. OSUtils::jsonDump(config["capabilities"], -1),
  1358. OSUtils::jsonString(config["identity"], ""),
  1359. (uint64_t)config["lastAuthorizedTime"],
  1360. (uint64_t)config["lastDeauthorizedTime"],
  1361. (bool)config["noAutoAssignIps"],
  1362. (int)config["remoteTraceLevel"],
  1363. target,
  1364. (uint64_t)config["revision"],
  1365. OSUtils::jsonDump(config["tags"], -1),
  1366. (int)config["vMajor"],
  1367. (int)config["vMinor"],
  1368. (int)config["vRev"],
  1369. (int)config["vProto"]);
  1370. }
  1371. if (! isNewMember) {
  1372. pqxx::result res = w.exec_params0("DELETE FROM ztc_member_ip_assignment WHERE member_id = $1 AND network_id = $2", memberId, networkId);
  1373. }
  1374. std::vector<std::string> assignments;
  1375. bool ipAssignError = false;
  1376. for (auto i = config["ipAssignments"].begin(); i != config["ipAssignments"].end(); ++i) {
  1377. std::string addr = *i;
  1378. if (std::find(assignments.begin(), assignments.end(), addr) != assignments.end()) {
  1379. continue;
  1380. }
  1381. pqxx::result res = w.exec_params0("INSERT INTO ztc_member_ip_assignment (member_id, network_id, address) VALUES ($1, $2, $3) ON CONFLICT (network_id, member_id, address) DO NOTHING", memberId, networkId, addr);
  1382. assignments.push_back(addr);
  1383. }
  1384. if (ipAssignError) {
  1385. fprintf(stderr, "%s: ipAssignError\n", _myAddressStr.c_str());
  1386. w.abort();
  1387. _pool->unborrow(c);
  1388. c.reset();
  1389. continue;
  1390. }
  1391. w.commit();
  1392. if (_smee != NULL && isNewMember) {
  1393. pqxx::row row = w.exec_params1(
  1394. "SELECT "
  1395. " count(h.hook_id) "
  1396. "FROM "
  1397. " ztc_hook h "
  1398. " INNER JOIN ztc_org o ON o.org_id = h.org_id "
  1399. " INNER JOIN ztc_network n ON n.owner_id = o.owner_id "
  1400. " WHERE "
  1401. "n.id = $1 ",
  1402. networkId);
  1403. int64_t hookCount = row[0].as<int64_t>();
  1404. if (hookCount > 0) {
  1405. notifyNewMember(networkId, memberId);
  1406. }
  1407. }
  1408. const uint64_t nwidInt = OSUtils::jsonIntHex(config["nwid"], 0ULL);
  1409. const uint64_t memberidInt = OSUtils::jsonIntHex(config["id"], 0ULL);
  1410. if (nwidInt && memberidInt) {
  1411. nlohmann::json nwOrig;
  1412. nlohmann::json memOrig;
  1413. nlohmann::json memNew(config);
  1414. get(nwidInt, nwOrig, memberidInt, memOrig);
  1415. _memberChanged(memOrig, memNew, qitem.second);
  1416. }
  1417. else {
  1418. fprintf(stderr, "%s: Can't notify of change. Error parsing nwid or memberid: %llu-%llu\n", _myAddressStr.c_str(), (unsigned long long)nwidInt, (unsigned long long)memberidInt);
  1419. }
  1420. }
  1421. catch (std::exception& e) {
  1422. fprintf(stderr, "%s ERROR: Error updating member %s-%s: %s\n", _myAddressStr.c_str(), networkId.c_str(), memberId.c_str(), e.what());
  1423. mspan->SetStatus(opentelemetry::trace::StatusCode::kError, e.what());
  1424. }
  1425. }
  1426. else if (objtype == "network") {
  1427. auto nspan = tracer->StartSpan("cv1::commitThread::network");
  1428. auto nscope = tracer->WithActiveSpan(nspan);
  1429. try {
  1430. // fprintf(stderr, "%s: commitThread: network\n", _myAddressStr.c_str());
  1431. pqxx::work w(*c->c);
  1432. std::string id = config["id"];
  1433. std::string remoteTraceTarget = "";
  1434. if (! config["remoteTraceTarget"].is_null()) {
  1435. remoteTraceTarget = config["remoteTraceTarget"];
  1436. }
  1437. std::string rulesSource = "";
  1438. if (config["rulesSource"].is_string()) {
  1439. rulesSource = config["rulesSource"];
  1440. }
  1441. // This ugly query exists because when we want to mirror networks to/from
  1442. // another data store (e.g. FileDB or LFDB) it is possible to get a network
  1443. // that doesn't exist in Central's database. This does an upsert and sets
  1444. // the owner_id to the "first" global admin in the user DB if the record
  1445. // did not previously exist. If the record already exists owner_id is left
  1446. // unchanged, so owner_id should be left out of the update clause.
  1447. pqxx::result res = w.exec_params0(
  1448. "INSERT INTO ztc_network (id, creation_time, owner_id, controller_id, capabilities, enable_broadcast, "
  1449. "last_modified, mtu, multicast_limit, name, private, "
  1450. "remote_trace_level, remote_trace_target, rules, rules_source, "
  1451. "tags, v4_assign_mode, v6_assign_mode, sso_enabled) VALUES ("
  1452. "$1, TO_TIMESTAMP($5::double precision/1000), "
  1453. "(SELECT user_id AS owner_id FROM ztc_global_permissions WHERE authorize = true AND del = true AND modify = true AND read = true LIMIT 1),"
  1454. "$2, $3, $4, TO_TIMESTAMP($5::double precision/1000), "
  1455. "$6, $7, $8, $9, $10, $11, $12, $13, $14, $15, $16, $17) "
  1456. "ON CONFLICT (id) DO UPDATE set controller_id = EXCLUDED.controller_id, "
  1457. "capabilities = EXCLUDED.capabilities, enable_broadcast = EXCLUDED.enable_broadcast, "
  1458. "last_modified = EXCLUDED.last_modified, mtu = EXCLUDED.mtu, "
  1459. "multicast_limit = EXCLUDED.multicast_limit, name = EXCLUDED.name, "
  1460. "private = EXCLUDED.private, remote_trace_level = EXCLUDED.remote_trace_level, "
  1461. "remote_trace_target = EXCLUDED.remote_trace_target, rules = EXCLUDED.rules, "
  1462. "rules_source = EXCLUDED.rules_source, tags = EXCLUDED.tags, "
  1463. "v4_assign_mode = EXCLUDED.v4_assign_mode, v6_assign_mode = EXCLUDED.v6_assign_mode, "
  1464. "sso_enabled = EXCLUDED.sso_enabled",
  1465. id,
  1466. _myAddressStr,
  1467. OSUtils::jsonDump(config["capabilities"], -1),
  1468. (bool)config["enableBroadcast"],
  1469. OSUtils::now(),
  1470. (int)config["mtu"],
  1471. (int)config["multicastLimit"],
  1472. OSUtils::jsonString(config["name"], ""),
  1473. (bool)config["private"],
  1474. (int)config["remoteTraceLevel"],
  1475. remoteTraceTarget,
  1476. OSUtils::jsonDump(config["rules"], -1),
  1477. rulesSource,
  1478. OSUtils::jsonDump(config["tags"], -1),
  1479. OSUtils::jsonDump(config["v4AssignMode"], -1),
  1480. OSUtils::jsonDump(config["v6AssignMode"], -1),
  1481. OSUtils::jsonBool(config["ssoEnabled"], false));
  1482. res = w.exec_params0("DELETE FROM ztc_network_assignment_pool WHERE network_id = $1", 0);
  1483. auto pool = config["ipAssignmentPools"];
  1484. bool err = false;
  1485. for (auto i = pool.begin(); i != pool.end(); ++i) {
  1486. std::string start = (*i)["ipRangeStart"];
  1487. std::string end = (*i)["ipRangeEnd"];
  1488. res = w.exec_params0(
  1489. "INSERT INTO ztc_network_assignment_pool (network_id, ip_range_start, ip_range_end) "
  1490. "VALUES ($1, $2, $3)",
  1491. id,
  1492. start,
  1493. end);
  1494. }
  1495. res = w.exec_params0("DELETE FROM ztc_network_route WHERE network_id = $1", id);
  1496. auto routes = config["routes"];
  1497. err = false;
  1498. for (auto i = routes.begin(); i != routes.end(); ++i) {
  1499. std::string t = (*i)["target"];
  1500. std::vector<std::string> target;
  1501. std::istringstream f(t);
  1502. std::string s;
  1503. while (std::getline(f, s, '/')) {
  1504. target.push_back(s);
  1505. }
  1506. if (target.empty() || target.size() != 2) {
  1507. continue;
  1508. }
  1509. std::string targetAddr = target[0];
  1510. std::string targetBits = target[1];
  1511. std::string via = "NULL";
  1512. if (! (*i)["via"].is_null()) {
  1513. via = (*i)["via"];
  1514. }
  1515. res = w.exec_params0("INSERT INTO ztc_network_route (network_id, address, bits, via) VALUES ($1, $2, $3, $4)", id, targetAddr, targetBits, (via == "NULL" ? NULL : via.c_str()));
  1516. }
  1517. if (err) {
  1518. fprintf(stderr, "%s: route add error\n", _myAddressStr.c_str());
  1519. w.abort();
  1520. _pool->unborrow(c);
  1521. continue;
  1522. }
  1523. auto dns = config["dns"];
  1524. std::string domain = dns["domain"];
  1525. std::stringstream servers;
  1526. servers << "{";
  1527. for (auto j = dns["servers"].begin(); j < dns["servers"].end(); ++j) {
  1528. servers << *j;
  1529. if ((j + 1) != dns["servers"].end()) {
  1530. servers << ",";
  1531. }
  1532. }
  1533. servers << "}";
  1534. std::string s = servers.str();
  1535. res = w.exec_params0("INSERT INTO ztc_network_dns (network_id, domain, servers) VALUES ($1, $2, $3) ON CONFLICT (network_id) DO UPDATE SET domain = EXCLUDED.domain, servers = EXCLUDED.servers", id, domain, s);
  1536. w.commit();
  1537. const uint64_t nwidInt = OSUtils::jsonIntHex(config["nwid"], 0ULL);
  1538. if (nwidInt) {
  1539. nlohmann::json nwOrig;
  1540. nlohmann::json nwNew(config);
  1541. get(nwidInt, nwOrig);
  1542. _networkChanged(nwOrig, nwNew, qitem.second);
  1543. }
  1544. else {
  1545. fprintf(stderr, "%s: Can't notify network changed: %llu\n", _myAddressStr.c_str(), (unsigned long long)nwidInt);
  1546. }
  1547. }
  1548. catch (std::exception& e) {
  1549. nspan->SetStatus(opentelemetry::trace::StatusCode::kError, e.what());
  1550. fprintf(stderr, "%s ERROR: Error updating network: %s\n", _myAddressStr.c_str(), e.what());
  1551. }
  1552. if (_redisMemberStatus) {
  1553. try {
  1554. std::string id = config["id"];
  1555. std::string controllerId = _myAddressStr.c_str();
  1556. std::string key = "networks:{" + controllerId + "}";
  1557. if (_rc->clusterMode) {
  1558. _cluster->sadd(key, id);
  1559. }
  1560. else {
  1561. _redis->sadd(key, id);
  1562. }
  1563. }
  1564. catch (sw::redis::Error& e) {
  1565. nspan->SetStatus(opentelemetry::trace::StatusCode::kError, e.what());
  1566. fprintf(stderr, "ERROR: Error adding network to Redis: %s\n", e.what());
  1567. }
  1568. }
  1569. }
  1570. else if (objtype == "_delete_network") {
  1571. auto dspan = tracer->StartSpan("cv1::commitThread::_delete_network");
  1572. auto dscope = tracer->WithActiveSpan(dspan);
  1573. // fprintf(stderr, "%s: commitThread: delete network\n", _myAddressStr.c_str());
  1574. try {
  1575. pqxx::work w(*c->c);
  1576. std::string networkId = config["nwid"];
  1577. pqxx::result res = w.exec_params0("UPDATE ztc_network SET deleted = true WHERE id = $1", networkId);
  1578. w.commit();
  1579. }
  1580. catch (std::exception& e) {
  1581. dspan->SetStatus(opentelemetry::trace::StatusCode::kError, e.what());
  1582. fprintf(stderr, "%s ERROR: Error deleting network: %s\n", _myAddressStr.c_str(), e.what());
  1583. }
  1584. if (_redisMemberStatus) {
  1585. try {
  1586. std::string id = config["id"];
  1587. std::string controllerId = _myAddressStr.c_str();
  1588. std::string key = "networks:{" + controllerId + "}";
  1589. if (_rc->clusterMode) {
  1590. _cluster->srem(key, id);
  1591. _cluster->del("network-nodes-online:{" + controllerId + "}:" + id);
  1592. }
  1593. else {
  1594. _redis->srem(key, id);
  1595. _redis->del("network-nodes-online:{" + controllerId + "}:" + id);
  1596. }
  1597. }
  1598. catch (sw::redis::Error& e) {
  1599. dspan->SetStatus(opentelemetry::trace::StatusCode::kError, e.what());
  1600. fprintf(stderr, "ERROR: Error adding network to Redis: %s\n", e.what());
  1601. }
  1602. }
  1603. }
  1604. else if (objtype == "_delete_member") {
  1605. auto mspan = tracer->StartSpan("cv1::commitThread::_delete_member");
  1606. auto mscope = tracer->WithActiveSpan(mspan);
  1607. // fprintf(stderr, "%s commitThread: delete member\n", _myAddressStr.c_str());
  1608. try {
  1609. pqxx::work w(*c->c);
  1610. std::string memberId = config["id"];
  1611. std::string networkId = config["nwid"];
  1612. pqxx::result res = w.exec_params0("UPDATE ztc_member SET hidden = true, deleted = true WHERE id = $1 AND network_id = $2", memberId, networkId);
  1613. w.commit();
  1614. }
  1615. catch (std::exception& e) {
  1616. mspan->SetStatus(opentelemetry::trace::StatusCode::kError, e.what());
  1617. fprintf(stderr, "%s ERROR: Error deleting member: %s\n", _myAddressStr.c_str(), e.what());
  1618. }
  1619. if (_redisMemberStatus) {
  1620. try {
  1621. std::string memberId = config["id"];
  1622. std::string networkId = config["nwid"];
  1623. std::string controllerId = _myAddressStr.c_str();
  1624. std::string key = "network-nodes-all:{" + controllerId + "}:" + networkId;
  1625. if (_rc->clusterMode) {
  1626. _cluster->srem(key, memberId);
  1627. _cluster->del("member:{" + controllerId + "}:" + networkId + ":" + memberId);
  1628. }
  1629. else {
  1630. _redis->srem(key, memberId);
  1631. _redis->del("member:{" + controllerId + "}:" + networkId + ":" + memberId);
  1632. }
  1633. }
  1634. catch (sw::redis::Error& e) {
  1635. mspan->SetStatus(opentelemetry::trace::StatusCode::kError, e.what());
  1636. fprintf(stderr, "ERROR: Error deleting member from Redis: %s\n", e.what());
  1637. }
  1638. }
  1639. }
  1640. else {
  1641. fprintf(stderr, "%s ERROR: unknown objtype\n", _myAddressStr.c_str());
  1642. }
  1643. }
  1644. catch (std::exception& e) {
  1645. span->SetStatus(opentelemetry::trace::StatusCode::kError, e.what());
  1646. fprintf(stderr, "%s ERROR: Error getting objtype: %s\n", _myAddressStr.c_str(), e.what());
  1647. }
  1648. _pool->unborrow(c);
  1649. c.reset();
  1650. }
  1651. fprintf(stderr, "%s commitThread finished\n", _myAddressStr.c_str());
  1652. }
  1653. void CV1::notifyNewMember(const std::string& networkID, const std::string& memberID)
  1654. {
  1655. auto provider = opentelemetry::trace::Provider::GetTracerProvider();
  1656. auto tracer = provider->GetTracer("cv1");
  1657. auto span = tracer->StartSpan("cv1::notifyNewMember");
  1658. auto scope = tracer->WithActiveSpan(span);
  1659. smeeclient::smee_client_notify_network_joined(_smee, networkID.c_str(), memberID.c_str());
  1660. }
  1661. void CV1::onlineNotificationThread()
  1662. {
  1663. waitForReady();
  1664. if (_redisMemberStatus) {
  1665. onlineNotification_Redis();
  1666. }
  1667. else {
  1668. onlineNotification_Postgres();
  1669. }
  1670. }
  1671. /**
  1672. * ONLY UNCOMMENT FOR TEMPORARY DB MAINTENANCE
  1673. *
  1674. * This define temporarily turns off writing to the member status table
  1675. * so it can be reindexed when the indexes get too large.
  1676. */
  1677. // #define DISABLE_MEMBER_STATUS 1
  1678. void CV1::onlineNotification_Postgres()
  1679. {
  1680. _connected = 1;
  1681. nlohmann::json jtmp1, jtmp2;
  1682. while (_run == 1) {
  1683. auto provider = opentelemetry::trace::Provider::GetTracerProvider();
  1684. auto tracer = provider->GetTracer("cv1");
  1685. auto span = tracer->StartSpan("cv1::onlineNotification_Postgres");
  1686. auto scope = tracer->WithActiveSpan(span);
  1687. auto c = _pool->borrow();
  1688. auto c2 = _pool->borrow();
  1689. try {
  1690. fprintf(stderr, "%s onlineNotification_Postgres\n", _myAddressStr.c_str());
  1691. std::unordered_map<std::pair<uint64_t, uint64_t>, NodeOnlineRecord, _PairHasher> lastOnline;
  1692. {
  1693. std::lock_guard<std::mutex> l(_lastOnline_l);
  1694. lastOnline.swap(_lastOnline);
  1695. }
  1696. #ifndef DISABLE_MEMBER_STATUS
  1697. pqxx::work w(*c->c);
  1698. pqxx::work w2(*c2->c);
  1699. fprintf(stderr, "online notification tick\n");
  1700. bool firstRun = true;
  1701. bool memberAdded = false;
  1702. int updateCount = 0;
  1703. pqxx::pipeline pipe(w);
  1704. for (auto i = lastOnline.begin(); i != lastOnline.end(); ++i) {
  1705. updateCount += 1;
  1706. uint64_t nwid_i = i->first.first;
  1707. char nwidTmp[64];
  1708. char memTmp[64];
  1709. char ipTmp[64];
  1710. OSUtils::ztsnprintf(nwidTmp, sizeof(nwidTmp), "%.16llx", nwid_i);
  1711. OSUtils::ztsnprintf(memTmp, sizeof(memTmp), "%.10llx", i->first.second);
  1712. if (! get(nwid_i, jtmp1, i->first.second, jtmp2)) {
  1713. continue; // skip non existent networks/members
  1714. }
  1715. std::string networkId(nwidTmp);
  1716. std::string memberId(memTmp);
  1717. try {
  1718. pqxx::row r = w2.exec_params1("SELECT id, network_id FROM ztc_member WHERE network_id = $1 AND id = $2", networkId, memberId);
  1719. }
  1720. catch (pqxx::unexpected_rows& e) {
  1721. continue;
  1722. }
  1723. int64_t ts = i->second.lastSeen;
  1724. std::string ipAddr = i->second.physicalAddress.toIpString(ipTmp);
  1725. std::string timestamp = std::to_string(ts);
  1726. std::string osArch = i->second.osArch;
  1727. std::vector<std::string> osArchSplit = split(osArch, '/');
  1728. std::string os = "unknown";
  1729. std::string arch = "unknown";
  1730. if (osArchSplit.size() == 2) {
  1731. os = osArchSplit[0];
  1732. arch = osArchSplit[1];
  1733. }
  1734. std::stringstream memberUpdate;
  1735. memberUpdate << "INSERT INTO ztc_member_status (network_id, member_id, address, last_updated, os, arch) VALUES "
  1736. << "('" << networkId << "', '" << memberId << "', ";
  1737. if (ipAddr.empty()) {
  1738. memberUpdate << "NULL, ";
  1739. }
  1740. else {
  1741. memberUpdate << "'" << ipAddr << "', ";
  1742. }
  1743. memberUpdate << "TO_TIMESTAMP(" << timestamp << "::double precision/1000), "
  1744. << "'" << os << "', "
  1745. << "'" << arch << "'"
  1746. << ") "
  1747. << " ON CONFLICT (network_id, member_id) DO UPDATE SET address = EXCLUDED.address, last_updated = EXCLUDED.last_updated, "
  1748. << "os = EXCLUDED.os, arch = EXCLUDED.arch";
  1749. pipe.insert(memberUpdate.str());
  1750. Metrics::pgsql_node_checkin++;
  1751. }
  1752. while (! pipe.empty()) {
  1753. pipe.retrieve();
  1754. }
  1755. pipe.complete();
  1756. w.commit();
  1757. fprintf(stderr, "%s: Updated online status of %d members\n", _myAddressStr.c_str(), updateCount);
  1758. #endif
  1759. }
  1760. catch (std::exception& e) {
  1761. fprintf(stderr, "%s: error in onlinenotification thread: %s\n", _myAddressStr.c_str(), e.what());
  1762. }
  1763. _pool->unborrow(c2);
  1764. _pool->unborrow(c);
  1765. ConnectionPoolStats stats = _pool->get_stats();
  1766. fprintf(stderr, "%s pool stats: in use size: %llu, available size: %llu, total: %llu\n", _myAddressStr.c_str(), stats.borrowed_size, stats.pool_size, (stats.borrowed_size + stats.pool_size));
  1767. span->End();
  1768. std::this_thread::sleep_for(std::chrono::seconds(10));
  1769. }
  1770. fprintf(stderr, "%s: Fell out of run loop in onlineNotificationThread\n", _myAddressStr.c_str());
  1771. if (_run == 1) {
  1772. fprintf(stderr, "ERROR: %s onlineNotificationThread should still be running! Exiting Controller.\n", _myAddressStr.c_str());
  1773. exit(6);
  1774. }
  1775. }
  1776. void CV1::onlineNotification_Redis()
  1777. {
  1778. _connected = 1;
  1779. char buf[11] = { 0 };
  1780. std::string controllerId = std::string(_myAddress.toString(buf));
  1781. while (_run == 1) {
  1782. auto provider = opentelemetry::trace::Provider::GetTracerProvider();
  1783. auto tracer = provider->GetTracer("cv1");
  1784. auto span = tracer->StartSpan("cv1::onlineNotification_Redis");
  1785. auto scope = tracer->WithActiveSpan(span);
  1786. fprintf(stderr, "onlineNotification tick\n");
  1787. auto start = std::chrono::high_resolution_clock::now();
  1788. uint64_t count = 0;
  1789. std::unordered_map<std::pair<uint64_t, uint64_t>, NodeOnlineRecord, _PairHasher> lastOnline;
  1790. {
  1791. std::lock_guard<std::mutex> l(_lastOnline_l);
  1792. lastOnline.swap(_lastOnline);
  1793. }
  1794. try {
  1795. if (! lastOnline.empty()) {
  1796. if (_rc->clusterMode) {
  1797. auto tx = _cluster->transaction(controllerId, true, false);
  1798. count = _doRedisUpdate(tx, controllerId, lastOnline);
  1799. }
  1800. else {
  1801. auto tx = _redis->transaction(true, false);
  1802. count = _doRedisUpdate(tx, controllerId, lastOnline);
  1803. }
  1804. }
  1805. }
  1806. catch (sw::redis::Error& e) {
  1807. fprintf(stderr, "Error in online notification thread (redis): %s\n", e.what());
  1808. }
  1809. auto end = std::chrono::high_resolution_clock::now();
  1810. auto dur = std::chrono::duration_cast<std::chrono::milliseconds>(end - start);
  1811. auto total = dur.count();
  1812. fprintf(stderr, "onlineNotification ran in %llu ms\n", total);
  1813. span->End();
  1814. std::this_thread::sleep_for(std::chrono::seconds(5));
  1815. }
  1816. }
  1817. uint64_t CV1::_doRedisUpdate(sw::redis::Transaction& tx, std::string& controllerId, std::unordered_map<std::pair<uint64_t, uint64_t>, NodeOnlineRecord, _PairHasher>& lastOnline)
  1818. {
  1819. auto provider = opentelemetry::trace::Provider::GetTracerProvider();
  1820. auto tracer = provider->GetTracer("cv1");
  1821. auto span = tracer->StartSpan("cv1::_doRedisUpdate");
  1822. auto scope = tracer->WithActiveSpan(span);
  1823. nlohmann::json jtmp1, jtmp2;
  1824. uint64_t count = 0;
  1825. for (auto i = lastOnline.begin(); i != lastOnline.end(); ++i) {
  1826. uint64_t nwid_i = i->first.first;
  1827. uint64_t memberid_i = i->first.second;
  1828. char nwidTmp[64];
  1829. char memTmp[64];
  1830. char ipTmp[64];
  1831. OSUtils::ztsnprintf(nwidTmp, sizeof(nwidTmp), "%.16llx", nwid_i);
  1832. OSUtils::ztsnprintf(memTmp, sizeof(memTmp), "%.10llx", memberid_i);
  1833. if (! get(nwid_i, jtmp1, memberid_i, jtmp2)) {
  1834. continue; // skip non existent members/networks
  1835. }
  1836. std::string networkId(nwidTmp);
  1837. std::string memberId(memTmp);
  1838. int64_t ts = i->second.lastSeen;
  1839. std::string ipAddr = i->second.physicalAddress.toIpString(ipTmp);
  1840. std::string timestamp = std::to_string(ts);
  1841. std::string osArch = i->second.osArch;
  1842. std::vector<std::string> osArchSplit = split(osArch, '/');
  1843. std::string os = "unknown";
  1844. std::string arch = "unknown";
  1845. if (osArchSplit.size() == 2) {
  1846. os = osArchSplit[0];
  1847. arch = osArchSplit[1];
  1848. }
  1849. std::unordered_map<std::string, std::string> record = { { "id", memberId }, { "address", ipAddr }, { "last_updated", std::to_string(ts) }, { "os", os }, { "arch", arch } };
  1850. tx.zadd("nodes-online:{" + controllerId + "}", memberId, ts)
  1851. .zadd("nodes-online2:{" + controllerId + "}", networkId + "-" + memberId, ts)
  1852. .zadd("network-nodes-online:{" + controllerId + "}:" + networkId, memberId, ts)
  1853. .zadd("active-networks:{" + controllerId + "}", networkId, ts)
  1854. .sadd("network-nodes-all:{" + controllerId + "}:" + networkId, memberId)
  1855. .hmset("member:{" + controllerId + "}:" + networkId + ":" + memberId, record.begin(), record.end());
  1856. ++count;
  1857. Metrics::redis_node_checkin++;
  1858. }
  1859. // expire records from all-nodes and network-nodes member list
  1860. uint64_t expireOld = OSUtils::now() - 300000;
  1861. tx.zremrangebyscore("nodes-online:{" + controllerId + "}", sw::redis::RightBoundedInterval<double>(expireOld, sw::redis::BoundType::LEFT_OPEN));
  1862. tx.zremrangebyscore("nodes-online2:{" + controllerId + "}", sw::redis::RightBoundedInterval<double>(expireOld, sw::redis::BoundType::LEFT_OPEN));
  1863. tx.zremrangebyscore("active-networks:{" + controllerId + "}", sw::redis::RightBoundedInterval<double>(expireOld, sw::redis::BoundType::LEFT_OPEN));
  1864. {
  1865. std::shared_lock<std::shared_mutex> l(_networks_l);
  1866. for (const auto& it : _networks) {
  1867. uint64_t nwid_i = it.first;
  1868. char nwidTmp[64];
  1869. OSUtils::ztsnprintf(nwidTmp, sizeof(nwidTmp), "%.16llx", nwid_i);
  1870. tx.zremrangebyscore("network-nodes-online:{" + controllerId + "}:" + nwidTmp, sw::redis::RightBoundedInterval<double>(expireOld, sw::redis::BoundType::LEFT_OPEN));
  1871. }
  1872. }
  1873. tx.exec();
  1874. fprintf(stderr, "%s: Updated online status of %d members\n", _myAddressStr.c_str(), count);
  1875. return count;
  1876. }
  1877. #endif // ZT_CONTROLLER_USE_LIBPQ