CV1.cpp 69 KB

1234567891011121314151617181920212223242526272829303132333435363738394041424344454647484950515253545556575859606162636465666768697071727374757677787980818283848586878889909192939495969798991001011021031041051061071081091101111121131141151161171181191201211221231241251261271281291301311321331341351361371381391401411421431441451461471481491501511521531541551561571581591601611621631641651661671681691701711721731741751761771781791801811821831841851861871881891901911921931941951961971981992002012022032042052062072082092102112122132142152162172182192202212222232242252262272282292302312322332342352362372382392402412422432442452462472482492502512522532542552562572582592602612622632642652662672682692702712722732742752762772782792802812822832842852862872882892902912922932942952962972982993003013023033043053063073083093103113123133143153163173183193203213223233243253263273283293303313323333343353363373383393403413423433443453463473483493503513523533543553563573583593603613623633643653663673683693703713723733743753763773783793803813823833843853863873883893903913923933943953963973983994004014024034044054064074084094104114124134144154164174184194204214224234244254264274284294304314324334344354364374384394404414424434444454464474484494504514524534544554564574584594604614624634644654664674684694704714724734744754764774784794804814824834844854864874884894904914924934944954964974984995005015025035045055065075085095105115125135145155165175185195205215225235245255265275285295305315325335345355365375385395405415425435445455465475485495505515525535545555565575585595605615625635645655665675685695705715725735745755765775785795805815825835845855865875885895905915925935945955965975985996006016026036046056066076086096106116126136146156166176186196206216226236246256266276286296306316326336346356366376386396406416426436446456466476486496506516526536546556566576586596606616626636646656666676686696706716726736746756766776786796806816826836846856866876886896906916926936946956966976986997007017027037047057067077087097107117127137147157167177187197207217227237247257267277287297307317327337347357367377387397407417427437447457467477487497507517527537547557567577587597607617627637647657667677687697707717727737747757767777787797807817827837847857867877887897907917927937947957967977987998008018028038048058068078088098108118128138148158168178188198208218228238248258268278288298308318328338348358368378388398408418428438448458468478488498508518528538548558568578588598608618628638648658668678688698708718728738748758768778788798808818828838848858868878888898908918928938948958968978988999009019029039049059069079089099109119129139149159169179189199209219229239249259269279289299309319329339349359369379389399409419429439449459469479489499509519529539549559569579589599609619629639649659669679689699709719729739749759769779789799809819829839849859869879889899909919929939949959969979989991000100110021003100410051006100710081009101010111012101310141015101610171018101910201021102210231024102510261027102810291030103110321033103410351036103710381039104010411042104310441045104610471048104910501051105210531054105510561057105810591060106110621063106410651066106710681069107010711072107310741075107610771078107910801081108210831084108510861087108810891090109110921093109410951096109710981099110011011102110311041105110611071108110911101111111211131114111511161117111811191120112111221123112411251126112711281129113011311132113311341135113611371138113911401141114211431144114511461147114811491150115111521153115411551156115711581159116011611162116311641165116611671168116911701171117211731174117511761177117811791180118111821183118411851186118711881189119011911192119311941195119611971198119912001201120212031204120512061207120812091210121112121213121412151216121712181219122012211222122312241225122612271228122912301231123212331234123512361237123812391240124112421243124412451246124712481249125012511252125312541255125612571258125912601261126212631264126512661267126812691270127112721273127412751276127712781279128012811282128312841285128612871288128912901291129212931294129512961297129812991300130113021303130413051306130713081309131013111312131313141315131613171318131913201321132213231324132513261327132813291330133113321333133413351336133713381339134013411342134313441345134613471348134913501351135213531354135513561357135813591360136113621363136413651366136713681369137013711372137313741375137613771378137913801381138213831384138513861387138813891390139113921393139413951396139713981399140014011402140314041405140614071408140914101411141214131414141514161417141814191420142114221423142414251426142714281429143014311432143314341435143614371438143914401441144214431444144514461447144814491450145114521453145414551456145714581459146014611462146314641465146614671468146914701471147214731474147514761477147814791480148114821483148414851486148714881489149014911492149314941495149614971498149915001501150215031504150515061507150815091510151115121513151415151516151715181519152015211522152315241525152615271528152915301531153215331534153515361537153815391540154115421543154415451546154715481549155015511552155315541555155615571558155915601561156215631564156515661567156815691570157115721573157415751576157715781579158015811582158315841585158615871588158915901591159215931594159515961597159815991600160116021603160416051606160716081609161016111612161316141615161616171618161916201621162216231624162516261627162816291630163116321633163416351636163716381639164016411642164316441645164616471648164916501651165216531654165516561657165816591660166116621663166416651666166716681669167016711672167316741675167616771678167916801681168216831684168516861687168816891690169116921693169416951696169716981699170017011702170317041705170617071708170917101711171217131714171517161717171817191720172117221723172417251726172717281729173017311732173317341735173617371738173917401741174217431744174517461747174817491750175117521753175417551756175717581759176017611762176317641765176617671768176917701771177217731774177517761777177817791780178117821783178417851786178717881789179017911792179317941795179617971798179918001801180218031804180518061807180818091810181118121813181418151816181718181819182018211822182318241825182618271828182918301831183218331834183518361837183818391840184118421843184418451846184718481849185018511852185318541855185618571858185918601861186218631864186518661867186818691870187118721873187418751876187718781879188018811882188318841885188618871888188918901891189218931894189518961897189818991900190119021903190419051906190719081909191019111912191319141915191619171918191919201921192219231924192519261927192819291930193119321933193419351936193719381939194019411942194319441945194619471948194919501951195219531954195519561957195819591960196119621963196419651966196719681969197019711972197319741975197619771978197919801981198219831984198519861987198819891990199119921993199419951996199719981999200020012002200320042005200620072008200920102011201220132014201520162017201820192020202120222023202420252026202720282029203020312032203320342035203620372038203920402041204220432044204520462047204820492050205120522053205420552056205720582059206020612062206320642065206620672068206920702071207220732074207520762077207820792080208120822083208420852086208720882089209020912092
  1. /*
  2. * Copyright (c)2019 ZeroTier, Inc.
  3. *
  4. * Use of this software is governed by the Business Source License included
  5. * in the LICENSE.TXT file in the project's root directory.
  6. *
  7. * Change Date: 2026-01-01
  8. *
  9. * On the date above, in accordance with the Business Source License, use
  10. * of this software will be governed by version 2.0 of the Apache License.
  11. */
  12. /****/
  13. #include "CV1.hpp"
  14. #ifdef ZT_CONTROLLER_USE_LIBPQ
  15. #include "../node/Constants.hpp"
  16. #include "../node/SHA512.hpp"
  17. #include "../version.h"
  18. #include "CtlUtil.hpp"
  19. #include "EmbeddedNetworkController.hpp"
  20. #include "Redis.hpp"
  21. #include "opentelemetry/trace/provider.h"
  22. #include <chrono>
  23. #include <climits>
  24. #include <iomanip>
  25. #include <libpq-fe.h>
  26. #include <smeeclient.h>
  27. #include <sstream>
  28. // #define REDIS_TRACE 1
  29. using json = nlohmann::json;
  30. namespace {
  31. static const int DB_MINIMUM_VERSION = 38;
  32. } // anonymous namespace
  33. using namespace ZeroTier;
  34. using Attrs = std::vector<std::pair<std::string, std::string> >;
  35. using Item = std::pair<std::string, Attrs>;
  36. using ItemStream = std::vector<Item>;
  37. CV1::CV1(const Identity& myId, const char* path, int listenPort, RedisConfig* rc)
  38. : DB()
  39. , _pool()
  40. , _myId(myId)
  41. , _myAddress(myId.address())
  42. , _ready(0)
  43. , _connected(1)
  44. , _run(1)
  45. , _waitNoticePrinted(false)
  46. , _listenPort(listenPort)
  47. , _rc(rc)
  48. , _redis(NULL)
  49. , _cluster(NULL)
  50. , _redisMemberStatus(false)
  51. , _smee(NULL)
  52. {
  53. auto provider = opentelemetry::trace::Provider::GetTracerProvider();
  54. auto tracer = provider->GetTracer("cv1");
  55. auto span = tracer->StartSpan("cv1::CV1");
  56. auto scope = tracer->WithActiveSpan(span);
  57. char myAddress[64];
  58. _myAddressStr = myId.address().toString(myAddress);
  59. _connString = std::string(path);
  60. auto f = std::make_shared<PostgresConnFactory>(_connString);
  61. _pool = std::make_shared<ConnectionPool<PostgresConnection> >(15, 5, std::static_pointer_cast<ConnectionFactory>(f));
  62. memset(_ssoPsk, 0, sizeof(_ssoPsk));
  63. char* const ssoPskHex = getenv("ZT_SSO_PSK");
  64. #ifdef ZT_TRACE
  65. fprintf(stderr, "ZT_SSO_PSK: %s\n", ssoPskHex);
  66. #endif
  67. if (ssoPskHex) {
  68. // SECURITY: note that ssoPskHex will always be null-terminated if libc actually
  69. // returns something non-NULL. If the hex encodes something shorter than 48 bytes,
  70. // it will be padded at the end with zeroes. If longer, it'll be truncated.
  71. Utils::unhex(ssoPskHex, _ssoPsk, sizeof(_ssoPsk));
  72. }
  73. const char* redisMemberStatus = getenv("ZT_REDIS_MEMBER_STATUS");
  74. if (redisMemberStatus && (strcmp(redisMemberStatus, "true") == 0)) {
  75. _redisMemberStatus = true;
  76. fprintf(stderr, "Using redis for member status\n");
  77. }
  78. auto c = _pool->borrow();
  79. pqxx::work txn { *c->c };
  80. pqxx::row r { txn.exec1("SELECT version FROM ztc_database") };
  81. int dbVersion = r[0].as<int>();
  82. txn.commit();
  83. if (dbVersion < DB_MINIMUM_VERSION) {
  84. 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);
  85. exit(1);
  86. }
  87. _pool->unborrow(c);
  88. if (_rc != NULL) {
  89. auto innerspan = tracer->StartSpan("cv1::CV1::configureRedis");
  90. auto innerscope = tracer->WithActiveSpan(innerspan);
  91. sw::redis::ConnectionOptions opts;
  92. sw::redis::ConnectionPoolOptions poolOpts;
  93. opts.host = _rc->hostname;
  94. opts.port = _rc->port;
  95. opts.password = _rc->password;
  96. opts.db = 0;
  97. opts.keep_alive = true;
  98. opts.connect_timeout = std::chrono::seconds(3);
  99. poolOpts.size = 25;
  100. poolOpts.wait_timeout = std::chrono::seconds(5);
  101. poolOpts.connection_lifetime = std::chrono::minutes(3);
  102. poolOpts.connection_idle_time = std::chrono::minutes(1);
  103. if (_rc->clusterMode) {
  104. innerspan->SetAttribute("cluster_mode", "true");
  105. fprintf(stderr, "Using Redis in Cluster Mode\n");
  106. _cluster = std::make_shared<sw::redis::RedisCluster>(opts, poolOpts);
  107. }
  108. else {
  109. innerspan->SetAttribute("cluster_mode", "false");
  110. fprintf(stderr, "Using Redis in Standalone Mode\n");
  111. _redis = std::make_shared<sw::redis::Redis>(opts, poolOpts);
  112. }
  113. }
  114. _readyLock.lock();
  115. fprintf(stderr, "[%s] NOTICE: %.10llx controller PostgreSQL waiting for initial data download..." ZT_EOL_S, ::_timestr(), (unsigned long long)_myAddress.toInt());
  116. _waitNoticePrinted = true;
  117. initializeNetworks();
  118. initializeMembers();
  119. _heartbeatThread = std::thread(&CV1::heartbeat, this);
  120. _membersDbWatcher = std::thread(&CV1::membersDbWatcher, this);
  121. _networksDbWatcher = std::thread(&CV1::networksDbWatcher, this);
  122. for (int i = 0; i < ZT_CENTRAL_CONTROLLER_COMMIT_THREADS; ++i) {
  123. _commitThread[i] = std::thread(&CV1::commitThread, this);
  124. }
  125. _onlineNotificationThread = std::thread(&CV1::onlineNotificationThread, this);
  126. configureSmee();
  127. }
  128. CV1::~CV1()
  129. {
  130. if (_smee != NULL) {
  131. smeeclient::smee_client_delete(_smee);
  132. _smee = NULL;
  133. }
  134. _run = 0;
  135. std::this_thread::sleep_for(std::chrono::milliseconds(100));
  136. _heartbeatThread.join();
  137. _membersDbWatcher.join();
  138. _networksDbWatcher.join();
  139. _commitQueue.stop();
  140. for (int i = 0; i < ZT_CENTRAL_CONTROLLER_COMMIT_THREADS; ++i) {
  141. _commitThread[i].join();
  142. }
  143. _onlineNotificationThread.join();
  144. }
  145. void CV1::configureSmee()
  146. {
  147. auto provider = opentelemetry::trace::Provider::GetTracerProvider();
  148. auto tracer = provider->GetTracer("cv1");
  149. auto span = tracer->StartSpan("cv1::configureSmee");
  150. auto scope = tracer->WithActiveSpan(span);
  151. const char* TEMPORAL_SCHEME = "ZT_TEMPORAL_SCHEME";
  152. const char* TEMPORAL_HOST = "ZT_TEMPORAL_HOST";
  153. const char* TEMPORAL_PORT = "ZT_TEMPORAL_PORT";
  154. const char* TEMPORAL_NAMESPACE = "ZT_TEMPORAL_NAMESPACE";
  155. const char* SMEE_TASK_QUEUE = "ZT_SMEE_TASK_QUEUE";
  156. const char* scheme = getenv(TEMPORAL_SCHEME);
  157. if (scheme == NULL) {
  158. scheme = "http";
  159. }
  160. const char* host = getenv(TEMPORAL_HOST);
  161. const char* port = getenv(TEMPORAL_PORT);
  162. const char* ns = getenv(TEMPORAL_NAMESPACE);
  163. const char* task_queue = getenv(SMEE_TASK_QUEUE);
  164. if (scheme != NULL && host != NULL && port != NULL && ns != NULL && task_queue != NULL) {
  165. fprintf(stderr, "creating smee client\n");
  166. std::string hostPort = std::string(scheme) + std::string("://") + std::string(host) + std::string(":") + std::string(port);
  167. this->_smee = smeeclient::smee_client_new(hostPort.c_str(), ns, task_queue);
  168. }
  169. else {
  170. fprintf(stderr, "Smee client not configured\n");
  171. }
  172. }
  173. bool CV1::waitForReady()
  174. {
  175. while (_ready < 2) {
  176. _readyLock.lock();
  177. _readyLock.unlock();
  178. }
  179. return true;
  180. }
  181. bool CV1::isReady()
  182. {
  183. return ((_ready == 2) && (_connected));
  184. }
  185. bool CV1::save(nlohmann::json& record, bool notifyListeners)
  186. {
  187. auto provider = opentelemetry::trace::Provider::GetTracerProvider();
  188. auto tracer = provider->GetTracer("cv1");
  189. auto span = tracer->StartSpan("cv1::save");
  190. auto scope = tracer->WithActiveSpan(span);
  191. bool modified = false;
  192. try {
  193. if (! record.is_object()) {
  194. fprintf(stderr, "record is not an object?!?\n");
  195. return false;
  196. }
  197. const std::string objtype = record["objtype"];
  198. if (objtype == "network") {
  199. // fprintf(stderr, "network save\n");
  200. const uint64_t nwid = OSUtils::jsonIntHex(record["id"], 0ULL);
  201. if (nwid) {
  202. nlohmann::json old;
  203. get(nwid, old);
  204. if ((! old.is_object()) || (! _compareRecords(old, record))) {
  205. record["revision"] = OSUtils::jsonInt(record["revision"], 0ULL) + 1ULL;
  206. _commitQueue.post(std::pair<nlohmann::json, bool>(record, notifyListeners));
  207. modified = true;
  208. }
  209. }
  210. }
  211. else if (objtype == "member") {
  212. std::string networkId = record["nwid"];
  213. std::string memberId = record["id"];
  214. const uint64_t nwid = OSUtils::jsonIntHex(record["nwid"], 0ULL);
  215. const uint64_t id = OSUtils::jsonIntHex(record["id"], 0ULL);
  216. // fprintf(stderr, "member save %s-%s\n", networkId.c_str(), memberId.c_str());
  217. if ((id) && (nwid)) {
  218. nlohmann::json network, old;
  219. get(nwid, network, id, old);
  220. if ((! old.is_object()) || (! _compareRecords(old, record))) {
  221. // fprintf(stderr, "commit queue post\n");
  222. record["revision"] = OSUtils::jsonInt(record["revision"], 0ULL) + 1ULL;
  223. _commitQueue.post(std::pair<nlohmann::json, bool>(record, notifyListeners));
  224. modified = true;
  225. }
  226. else {
  227. // fprintf(stderr, "no change\n");
  228. }
  229. }
  230. }
  231. else {
  232. fprintf(stderr, "uhh waaat\n");
  233. }
  234. }
  235. catch (std::exception& e) {
  236. fprintf(stderr, "Error on PostgreSQL::save: %s\n", e.what());
  237. }
  238. catch (...) {
  239. fprintf(stderr, "Unknown error on PostgreSQL::save\n");
  240. }
  241. return modified;
  242. }
  243. void CV1::eraseNetwork(const uint64_t networkId)
  244. {
  245. auto provider = opentelemetry::trace::Provider::GetTracerProvider();
  246. auto tracer = provider->GetTracer("cv1");
  247. auto span = tracer->StartSpan("cv1::eraseNetwork");
  248. auto scope = tracer->WithActiveSpan(span);
  249. char networkIdStr[17];
  250. span->SetAttribute("network_id", Utils::hex(networkId, networkIdStr));
  251. fprintf(stderr, "PostgreSQL::eraseNetwork\n");
  252. char tmp2[24];
  253. waitForReady();
  254. Utils::hex(networkId, tmp2);
  255. std::pair<nlohmann::json, bool> tmp;
  256. tmp.first["id"] = tmp2;
  257. tmp.first["objtype"] = "_delete_network";
  258. tmp.second = true;
  259. _commitQueue.post(tmp);
  260. nlohmann::json nullJson;
  261. _networkChanged(tmp.first, nullJson, true);
  262. }
  263. void CV1::eraseMember(const uint64_t networkId, const uint64_t memberId)
  264. {
  265. auto provider = opentelemetry::trace::Provider::GetTracerProvider();
  266. auto tracer = provider->GetTracer("cv1");
  267. auto span = tracer->StartSpan("cv1::eraseMember");
  268. auto scope = tracer->WithActiveSpan(span);
  269. char networkIdStr[17];
  270. char memberIdStr[11];
  271. span->SetAttribute("network_id", Utils::hex(networkId, networkIdStr));
  272. span->SetAttribute("member_id", Utils::hex10(memberId, memberIdStr));
  273. fprintf(stderr, "PostgreSQL::eraseMember\n");
  274. char tmp2[24];
  275. waitForReady();
  276. std::pair<nlohmann::json, bool> tmp, nw;
  277. Utils::hex(networkId, tmp2);
  278. tmp.first["nwid"] = tmp2;
  279. Utils::hex(memberId, tmp2);
  280. tmp.first["id"] = tmp2;
  281. tmp.first["objtype"] = "_delete_member";
  282. tmp.second = true;
  283. _commitQueue.post(tmp);
  284. nlohmann::json nullJson;
  285. _memberChanged(tmp.first, nullJson, true);
  286. }
  287. void CV1::nodeIsOnline(const uint64_t networkId, const uint64_t memberId, const InetAddress& physicalAddress, const char* osArch)
  288. {
  289. auto provider = opentelemetry::trace::Provider::GetTracerProvider();
  290. auto tracer = provider->GetTracer("cv1");
  291. auto span = tracer->StartSpan("cv1::nodeIsOnline");
  292. auto scope = tracer->WithActiveSpan(span);
  293. char networkIdStr[17];
  294. char memberIdStr[11];
  295. char ipStr[INET6_ADDRSTRLEN];
  296. span->SetAttribute("network_id", Utils::hex(networkId, networkIdStr));
  297. span->SetAttribute("member_id", Utils::hex10(memberId, memberIdStr));
  298. span->SetAttribute("physical_address", physicalAddress.toString(ipStr));
  299. span->SetAttribute("os_arch", osArch);
  300. std::lock_guard<std::mutex> l(_lastOnline_l);
  301. NodeOnlineRecord& i = _lastOnline[std::pair<uint64_t, uint64_t>(networkId, memberId)];
  302. i.lastSeen = OSUtils::now();
  303. if (physicalAddress) {
  304. i.physicalAddress = physicalAddress;
  305. }
  306. i.osArch = std::string(osArch);
  307. }
  308. void CV1::nodeIsOnline(const uint64_t networkId, const uint64_t memberId, const InetAddress& physicalAddress)
  309. {
  310. this->nodeIsOnline(networkId, memberId, physicalAddress, "unknown/unknown");
  311. }
  312. AuthInfo CV1::getSSOAuthInfo(const nlohmann::json& member, const std::string& redirectURL)
  313. {
  314. auto provider = opentelemetry::trace::Provider::GetTracerProvider();
  315. auto tracer = provider->GetTracer("cv1");
  316. auto span = tracer->StartSpan("cv1::getSSOAuthInfo");
  317. auto scope = tracer->WithActiveSpan(span);
  318. Metrics::db_get_sso_info++;
  319. // NONCE is just a random character string. no semantic meaning
  320. // state = HMAC SHA384 of Nonce based on shared sso key
  321. //
  322. // need nonce timeout in database? make sure it's used within X time
  323. // X is 5 minutes for now. Make configurable later?
  324. //
  325. // how do we tell when a nonce is used? if auth_expiration_time is set
  326. std::string networkId = member["nwid"];
  327. std::string memberId = member["id"];
  328. char authenticationURL[4096] = { 0 };
  329. AuthInfo info;
  330. info.enabled = true;
  331. // if (memberId == "a10dccea52" && networkId == "8056c2e21c24673d") {
  332. // fprintf(stderr, "invalid authinfo for grant's machine\n");
  333. // info.version=1;
  334. // return info;
  335. // }
  336. // fprintf(stderr, "PostgreSQL::updateMemberOnLoad: %s-%s\n", networkId.c_str(), memberId.c_str());
  337. std::shared_ptr<PostgresConnection> c;
  338. try {
  339. c = _pool->borrow();
  340. pqxx::work w(*c->c);
  341. char nonceBytes[16] = { 0 };
  342. std::string nonce = "";
  343. // check if the member exists first.
  344. pqxx::row count = w.exec_params1("SELECT count(id) FROM ztc_member WHERE id = $1 AND network_id = $2 AND deleted = false", memberId, networkId);
  345. if (count[0].as<int>() == 1) {
  346. // get active nonce, if 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 ((NOW() AT TIME ZONE 'UTC') <= authentication_expiry_time) AND ((NOW() AT TIME ZONE 'UTC') <= nonce_expiration)",
  351. networkId,
  352. memberId);
  353. if (r.size() == 0) {
  354. // no active nonce.
  355. // find an unused nonce, if one exists.
  356. pqxx::result r = w.exec_params(
  357. "SELECT nonce FROM ztc_sso_expiry "
  358. "WHERE network_id = $1 AND member_id = $2 "
  359. "AND authentication_expiry_time IS NULL AND ((NOW() AT TIME ZONE 'UTC') <= nonce_expiration)",
  360. networkId,
  361. memberId);
  362. if (r.size() == 1) {
  363. // we have an existing nonce. Use it
  364. nonce = r.at(0)[0].as<std::string>();
  365. Utils::unhex(nonce.c_str(), nonceBytes, sizeof(nonceBytes));
  366. }
  367. else if (r.empty()) {
  368. // create a nonce
  369. Utils::getSecureRandom(nonceBytes, 16);
  370. char nonceBuf[64] = { 0 };
  371. Utils::hex(nonceBytes, sizeof(nonceBytes), nonceBuf);
  372. nonce = std::string(nonceBuf);
  373. pqxx::result ir = w.exec_params0(
  374. "INSERT INTO ztc_sso_expiry "
  375. "(nonce, nonce_expiration, network_id, member_id) VALUES "
  376. "($1, TO_TIMESTAMP($2::double precision/1000), $3, $4)",
  377. nonce,
  378. OSUtils::now() + 300000,
  379. networkId,
  380. memberId);
  381. w.commit();
  382. }
  383. else {
  384. // > 1 ?!? Thats an error!
  385. fprintf(stderr, "> 1 unused nonce!\n");
  386. exit(6);
  387. }
  388. }
  389. else if (r.size() == 1) {
  390. nonce = r.at(0)[0].as<std::string>();
  391. Utils::unhex(nonce.c_str(), nonceBytes, sizeof(nonceBytes));
  392. }
  393. else {
  394. // more than 1 nonce in use? Uhhh...
  395. fprintf(stderr, "> 1 nonce in use for network member?!?\n");
  396. exit(7);
  397. }
  398. r = w.exec_params(
  399. "SELECT oc.client_id, oc.authorization_endpoint, oc.issuer, oc.provider, oc.sso_impl_version "
  400. "FROM ztc_network AS n "
  401. "INNER JOIN ztc_org o "
  402. " ON o.owner_id = n.owner_id "
  403. "LEFT OUTER JOIN ztc_network_oidc_config noc "
  404. " ON noc.network_id = n.id "
  405. "LEFT OUTER JOIN ztc_oidc_config oc "
  406. " ON noc.client_id = oc.client_id AND oc.org_id = o.org_id "
  407. "WHERE n.id = $1 AND n.sso_enabled = true",
  408. networkId);
  409. std::string client_id = "";
  410. std::string authorization_endpoint = "";
  411. std::string issuer = "";
  412. std::string provider = "";
  413. uint64_t sso_version = 0;
  414. if (r.size() == 1) {
  415. client_id = r.at(0)[0].as<std::optional<std::string> >().value_or("");
  416. authorization_endpoint = r.at(0)[1].as<std::optional<std::string> >().value_or("");
  417. issuer = r.at(0)[2].as<std::optional<std::string> >().value_or("");
  418. provider = r.at(0)[3].as<std::optional<std::string> >().value_or("");
  419. sso_version = r.at(0)[4].as<std::optional<uint64_t> >().value_or(1);
  420. }
  421. else if (r.size() > 1) {
  422. fprintf(stderr, "ERROR: More than one auth endpoint for an organization?!?!? NetworkID: %s\n", networkId.c_str());
  423. }
  424. else {
  425. fprintf(stderr, "No client or auth endpoint?!?\n");
  426. }
  427. info.version = sso_version;
  428. // no catch all else because we don't actually care if no records exist here. just continue as normal.
  429. if ((! client_id.empty()) && (! authorization_endpoint.empty())) {
  430. uint8_t state[48];
  431. HMACSHA384(_ssoPsk, nonceBytes, sizeof(nonceBytes), state);
  432. char state_hex[256];
  433. Utils::hex(state, 48, state_hex);
  434. if (info.version == 0) {
  435. char url[2048] = { 0 };
  436. OSUtils::ztsnprintf(
  437. url,
  438. sizeof(authenticationURL),
  439. "%s?response_type=id_token&response_mode=form_post&scope=openid+email+profile&redirect_uri=%s&nonce=%s&state=%s&client_id=%s",
  440. authorization_endpoint.c_str(),
  441. url_encode(redirectURL).c_str(),
  442. nonce.c_str(),
  443. state_hex,
  444. client_id.c_str());
  445. info.authenticationURL = std::string(url);
  446. }
  447. else if (info.version == 1) {
  448. info.ssoClientID = client_id;
  449. info.issuerURL = issuer;
  450. info.ssoProvider = provider;
  451. info.ssoNonce = nonce;
  452. info.ssoState = std::string(state_hex) + "_" + networkId;
  453. info.centralAuthURL = redirectURL;
  454. #ifdef ZT_DEBUG
  455. fprintf(
  456. stderr,
  457. "ssoClientID: %s\nissuerURL: %s\nssoNonce: %s\nssoState: %s\ncentralAuthURL: %s\nprovider: %s\n",
  458. info.ssoClientID.c_str(),
  459. info.issuerURL.c_str(),
  460. info.ssoNonce.c_str(),
  461. info.ssoState.c_str(),
  462. info.centralAuthURL.c_str(),
  463. provider.c_str());
  464. #endif
  465. }
  466. }
  467. else {
  468. fprintf(stderr, "client_id: %s\nauthorization_endpoint: %s\n", client_id.c_str(), authorization_endpoint.c_str());
  469. }
  470. }
  471. _pool->unborrow(c);
  472. }
  473. catch (std::exception& e) {
  474. span->SetStatus(opentelemetry::trace::StatusCode::kError, e.what());
  475. fprintf(stderr, "ERROR: Error updating member on load for network %s: %s\n", networkId.c_str(), e.what());
  476. }
  477. return info; // std::string(authenticationURL);
  478. }
  479. void CV1::initializeNetworks()
  480. {
  481. auto provider = opentelemetry::trace::Provider::GetTracerProvider();
  482. auto tracer = provider->GetTracer("cv1");
  483. auto span = tracer->StartSpan("cv1::initializeNetworks");
  484. auto scope = tracer->WithActiveSpan(span);
  485. try {
  486. std::string setKey = "networks:{" + _myAddressStr + "}";
  487. fprintf(stderr, "Initializing Networks...\n");
  488. if (_redisMemberStatus) {
  489. fprintf(stderr, "Init Redis for networks...\n");
  490. try {
  491. if (_rc->clusterMode) {
  492. _cluster->del(setKey);
  493. }
  494. else {
  495. _redis->del(setKey);
  496. }
  497. }
  498. catch (sw::redis::Error& e) {
  499. // ignore. if this key doesn't exist, there's no reason to delete it
  500. }
  501. }
  502. std::unordered_set<std::string> networkSet;
  503. char qbuf[2048] = { 0 };
  504. sprintf(
  505. qbuf,
  506. "SELECT n.id, (EXTRACT(EPOCH FROM n.creation_time AT TIME ZONE 'UTC')*1000)::bigint as creation_time, n.capabilities, "
  507. "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, "
  508. "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, "
  509. "(CASE WHEN n.sso_enabled THEN oc.authorization_endpoint ELSE NULL END) as authorization_endpoint, "
  510. "(CASE WHEN n.sso_enabled THEN oc.provider ELSE NULL END) as provider, d.domain, d.servers, "
  511. "ARRAY(SELECT CONCAT(host(ip_range_start),'|', host(ip_range_end)) FROM ztc_network_assignment_pool WHERE network_id = n.id) AS assignment_pool, "
  512. "ARRAY(SELECT CONCAT(host(address),'/',bits::text,'|',COALESCE(host(via), 'NULL'))FROM ztc_network_route WHERE network_id = n.id) AS routes "
  513. "FROM ztc_network n "
  514. "LEFT OUTER JOIN ztc_org o "
  515. " ON o.owner_id = n.owner_id "
  516. "LEFT OUTER JOIN ztc_network_oidc_config noc "
  517. " ON noc.network_id = n.id "
  518. "LEFT OUTER JOIN ztc_oidc_config oc "
  519. " ON noc.client_id = oc.client_id AND oc.org_id = o.org_id "
  520. "LEFT OUTER JOIN ztc_network_dns d "
  521. " ON d.network_id = n.id "
  522. "WHERE deleted = false AND controller_id = '%s'",
  523. _myAddressStr.c_str());
  524. auto c = _pool->borrow();
  525. auto c2 = _pool->borrow();
  526. pqxx::work w { *c->c };
  527. fprintf(stderr, "Load networks from psql...\n");
  528. auto stream = pqxx::stream_from::query(w, qbuf);
  529. std::tuple<
  530. std::string // network ID
  531. ,
  532. std::optional<int64_t> // creationTime
  533. ,
  534. std::optional<std::string> // capabilities
  535. ,
  536. std::optional<bool> // enableBroadcast
  537. ,
  538. std::optional<uint64_t> // lastModified
  539. ,
  540. std::optional<int> // mtu
  541. ,
  542. std::optional<int> // multicastLimit
  543. ,
  544. std::optional<std::string> // name
  545. ,
  546. bool // private
  547. ,
  548. std::optional<int> // remoteTraceLevel
  549. ,
  550. std::optional<std::string> // remoteTraceTarget
  551. ,
  552. std::optional<uint64_t> // revision
  553. ,
  554. std::optional<std::string> // rules
  555. ,
  556. std::optional<std::string> // tags
  557. ,
  558. std::optional<std::string> // v4AssignMode
  559. ,
  560. std::optional<std::string> // v6AssignMode
  561. ,
  562. std::optional<bool> // ssoEnabled
  563. ,
  564. std::optional<std::string> // clientId
  565. ,
  566. std::optional<std::string> // authorizationEndpoint
  567. ,
  568. std::optional<std::string> // ssoProvider
  569. ,
  570. std::optional<std::string> // domain
  571. ,
  572. std::optional<std::string> // servers
  573. ,
  574. std::string // assignmentPoolString
  575. ,
  576. std::string // routeString
  577. >
  578. row;
  579. uint64_t count = 0;
  580. auto tmp = std::chrono::high_resolution_clock::now();
  581. uint64_t total = 0;
  582. while (stream >> row) {
  583. auto start = std::chrono::high_resolution_clock::now();
  584. json empty;
  585. json config;
  586. initNetwork(config);
  587. std::string nwid = std::get<0>(row);
  588. std::optional<int64_t> creationTime = std::get<1>(row);
  589. std::optional<std::string> capabilities = std::get<2>(row);
  590. std::optional<bool> enableBroadcast = std::get<3>(row);
  591. std::optional<uint64_t> lastModified = std::get<4>(row);
  592. std::optional<int> mtu = std::get<5>(row);
  593. std::optional<int> multicastLimit = std::get<6>(row);
  594. std::optional<std::string> name = std::get<7>(row);
  595. bool isPrivate = std::get<8>(row);
  596. std::optional<int> remoteTraceLevel = std::get<9>(row);
  597. std::optional<std::string> remoteTraceTarget = std::get<10>(row);
  598. std::optional<uint64_t> revision = std::get<11>(row);
  599. std::optional<std::string> rules = std::get<12>(row);
  600. std::optional<std::string> tags = std::get<13>(row);
  601. std::optional<std::string> v4AssignMode = std::get<14>(row);
  602. std::optional<std::string> v6AssignMode = std::get<15>(row);
  603. std::optional<bool> ssoEnabled = std::get<16>(row);
  604. std::optional<std::string> clientId = std::get<17>(row);
  605. std::optional<std::string> authorizationEndpoint = std::get<18>(row);
  606. std::optional<std::string> ssoProvider = std::get<19>(row);
  607. std::optional<std::string> dnsDomain = std::get<20>(row);
  608. std::optional<std::string> dnsServers = std::get<21>(row);
  609. std::string assignmentPoolString = std::get<22>(row);
  610. std::string routesString = std::get<23>(row);
  611. config["id"] = nwid;
  612. config["nwid"] = nwid;
  613. config["creationTime"] = creationTime.value_or(0);
  614. config["capabilities"] = json::parse(capabilities.value_or("[]"));
  615. config["enableBroadcast"] = enableBroadcast.value_or(false);
  616. config["lastModified"] = lastModified.value_or(0);
  617. config["mtu"] = mtu.value_or(2800);
  618. config["multicastLimit"] = multicastLimit.value_or(64);
  619. config["name"] = name.value_or("");
  620. config["private"] = isPrivate;
  621. config["remoteTraceLevel"] = remoteTraceLevel.value_or(0);
  622. config["remoteTraceTarget"] = remoteTraceTarget.value_or("");
  623. config["revision"] = revision.value_or(0);
  624. config["rules"] = json::parse(rules.value_or("[]"));
  625. config["tags"] = json::parse(tags.value_or("[]"));
  626. config["v4AssignMode"] = json::parse(v4AssignMode.value_or("{}"));
  627. config["v6AssignMode"] = json::parse(v6AssignMode.value_or("{}"));
  628. config["ssoEnabled"] = ssoEnabled.value_or(false);
  629. config["objtype"] = "network";
  630. config["ipAssignmentPools"] = json::array();
  631. config["routes"] = json::array();
  632. config["clientId"] = clientId.value_or("");
  633. config["authorizationEndpoint"] = authorizationEndpoint.value_or("");
  634. config["provider"] = ssoProvider.value_or("");
  635. networkSet.insert(nwid);
  636. if (dnsDomain.has_value()) {
  637. std::string serverList = dnsServers.value();
  638. json obj;
  639. auto servers = json::array();
  640. if (serverList.rfind("{", 0) != std::string::npos) {
  641. serverList = serverList.substr(1, serverList.size() - 2);
  642. std::stringstream ss(serverList);
  643. while (ss.good()) {
  644. std::string server;
  645. std::getline(ss, server, ',');
  646. servers.push_back(server);
  647. }
  648. }
  649. obj["domain"] = dnsDomain.value();
  650. obj["servers"] = servers;
  651. config["dns"] = obj;
  652. }
  653. config["ipAssignmentPools"] = json::array();
  654. if (assignmentPoolString != "{}") {
  655. std::string tmp = assignmentPoolString.substr(1, assignmentPoolString.size() - 2);
  656. std::vector<std::string> assignmentPools = split(tmp, ',');
  657. for (auto it = assignmentPools.begin(); it != assignmentPools.end(); ++it) {
  658. std::vector<std::string> r = split(*it, '|');
  659. json ip;
  660. ip["ipRangeStart"] = r[0];
  661. ip["ipRangeEnd"] = r[1];
  662. config["ipAssignmentPools"].push_back(ip);
  663. }
  664. }
  665. config["routes"] = json::array();
  666. if (routesString != "{}") {
  667. std::string tmp = routesString.substr(1, routesString.size() - 2);
  668. std::vector<std::string> routes = split(tmp, ',');
  669. for (auto it = routes.begin(); it != routes.end(); ++it) {
  670. std::vector<std::string> r = split(*it, '|');
  671. json route;
  672. route["target"] = r[0];
  673. route["via"] = ((route["via"] == "NULL") ? nullptr : r[1]);
  674. config["routes"].push_back(route);
  675. }
  676. }
  677. Metrics::network_count++;
  678. _networkChanged(empty, config, false);
  679. auto end = std::chrono::high_resolution_clock::now();
  680. auto dur = std::chrono::duration_cast<std::chrono::microseconds>(end - start);
  681. ;
  682. total += dur.count();
  683. ++count;
  684. if (count > 0 && count % 10000 == 0) {
  685. fprintf(stderr, "Averaging %llu us per network\n", (total / count));
  686. }
  687. }
  688. if (count > 0) {
  689. fprintf(stderr, "Took %llu us per network to load\n", (total / count));
  690. }
  691. stream.complete();
  692. w.commit();
  693. _pool->unborrow(c2);
  694. _pool->unborrow(c);
  695. fprintf(stderr, "done.\n");
  696. if (! networkSet.empty()) {
  697. if (_redisMemberStatus) {
  698. fprintf(stderr, "adding networks to redis...\n");
  699. if (_rc->clusterMode) {
  700. auto tx = _cluster->transaction(_myAddressStr, true, false);
  701. uint64_t count = 0;
  702. for (std::string nwid : networkSet) {
  703. tx.sadd(setKey, nwid);
  704. if (++count % 30000 == 0) {
  705. tx.exec();
  706. tx = _cluster->transaction(_myAddressStr, true, false);
  707. }
  708. }
  709. tx.exec();
  710. }
  711. else {
  712. auto tx = _redis->transaction(true, false);
  713. uint64_t count = 0;
  714. for (std::string nwid : networkSet) {
  715. tx.sadd(setKey, nwid);
  716. if (++count % 30000 == 0) {
  717. tx.exec();
  718. tx = _redis->transaction(true, false);
  719. }
  720. }
  721. tx.exec();
  722. }
  723. fprintf(stderr, "done.\n");
  724. }
  725. }
  726. if (++this->_ready == 2) {
  727. if (_waitNoticePrinted) {
  728. fprintf(stderr, "[%s] NOTICE: %.10llx controller PostgreSQL data download complete." ZT_EOL_S, _timestr(), (unsigned long long)_myAddress.toInt());
  729. }
  730. _readyLock.unlock();
  731. }
  732. fprintf(stderr, "network init done.\n");
  733. }
  734. catch (sw::redis::Error& e) {
  735. span->SetStatus(opentelemetry::trace::StatusCode::kError, e.what());
  736. fprintf(stderr, "ERROR: Error initializing networks in Redis: %s\n", e.what());
  737. std::this_thread::sleep_for(std::chrono::milliseconds(5000));
  738. exit(-1);
  739. }
  740. catch (std::exception& e) {
  741. span->SetStatus(opentelemetry::trace::StatusCode::kError, e.what());
  742. fprintf(stderr, "ERROR: Error initializing networks: %s\n", e.what());
  743. std::this_thread::sleep_for(std::chrono::milliseconds(5000));
  744. exit(-1);
  745. }
  746. }
  747. void CV1::initializeMembers()
  748. {
  749. auto provider = opentelemetry::trace::Provider::GetTracerProvider();
  750. auto tracer = provider->GetTracer("cv1");
  751. auto span = tracer->StartSpan("cv1::initializeMembers");
  752. auto scope = tracer->WithActiveSpan(span);
  753. std::string memberId;
  754. std::string networkId;
  755. try {
  756. std::unordered_map<std::string, std::string> networkMembers;
  757. fprintf(stderr, "Initializing Members...\n");
  758. std::string setKeyBase = "network-nodes-all:{" + _myAddressStr + "}:";
  759. if (_redisMemberStatus) {
  760. fprintf(stderr, "Initialize Redis for members...\n");
  761. std::unique_lock<std::shared_mutex> l(_networks_l);
  762. std::unordered_set<std::string> deletes;
  763. for (auto it : _networks) {
  764. uint64_t nwid_i = it.first;
  765. char nwidTmp[64] = { 0 };
  766. OSUtils::ztsnprintf(nwidTmp, sizeof(nwidTmp), "%.16llx", nwid_i);
  767. std::string nwid(nwidTmp);
  768. std::string key = setKeyBase + nwid;
  769. deletes.insert(key);
  770. }
  771. if (! deletes.empty()) {
  772. try {
  773. if (_rc->clusterMode) {
  774. auto tx = _cluster->transaction(_myAddressStr, true, false);
  775. for (std::string k : deletes) {
  776. tx.del(k);
  777. }
  778. tx.exec();
  779. }
  780. else {
  781. auto tx = _redis->transaction(true, false);
  782. for (std::string k : deletes) {
  783. tx.del(k);
  784. }
  785. tx.exec();
  786. }
  787. }
  788. catch (sw::redis::Error& e) {
  789. // ignore
  790. }
  791. }
  792. }
  793. char qbuf[2048];
  794. sprintf(
  795. qbuf,
  796. "SELECT m.id, m.network_id, m.active_bridge, m.authorized, m.capabilities, "
  797. "(EXTRACT(EPOCH FROM m.creation_time AT TIME ZONE 'UTC')*1000)::bigint, m.identity, "
  798. "(EXTRACT(EPOCH FROM m.last_authorized_time AT TIME ZONE 'UTC')*1000)::bigint, "
  799. "(EXTRACT(EPOCH FROM m.last_deauthorized_time AT TIME ZONE 'UTC')*1000)::bigint, "
  800. "m.remote_trace_level, m.remote_trace_target, m.tags, m.v_major, m.v_minor, m.v_rev, m.v_proto, "
  801. "m.no_auto_assign_ips, m.revision, m.sso_exempt, "
  802. "(CASE WHEN n.sso_enabled = TRUE AND m.sso_exempt = FALSE THEN "
  803. " ( "
  804. " SELECT (EXTRACT(EPOCH FROM e.authentication_expiry_time)*1000)::bigint "
  805. " FROM ztc_sso_expiry e "
  806. " INNER JOIN ztc_network n1 "
  807. " ON n1.id = e.network_id AND n1.deleted = TRUE "
  808. " 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 "
  809. " ORDER BY e.authentication_expiry_time DESC LIMIT 1 "
  810. " ) "
  811. " ELSE NULL "
  812. " END) AS authentication_expiry_time, "
  813. "ARRAY(SELECT DISTINCT address FROM ztc_member_ip_assignment WHERE member_id = m.id AND network_id = m.network_id) AS assigned_addresses "
  814. "FROM ztc_member m "
  815. "INNER JOIN ztc_network n "
  816. " ON n.id = m.network_id "
  817. "WHERE n.controller_id = '%s' AND n.deleted = FALSE AND m.deleted = FALSE",
  818. _myAddressStr.c_str());
  819. auto c = _pool->borrow();
  820. auto c2 = _pool->borrow();
  821. pqxx::work w { *c->c };
  822. fprintf(stderr, "Load members from psql...\n");
  823. auto stream = pqxx::stream_from::query(w, qbuf);
  824. std::tuple<
  825. std::string // memberId
  826. ,
  827. std::string // memberId
  828. ,
  829. std::optional<bool> // activeBridge
  830. ,
  831. std::optional<bool> // authorized
  832. ,
  833. std::optional<std::string> // capabilities
  834. ,
  835. std::optional<uint64_t> // creationTime
  836. ,
  837. std::optional<std::string> // identity
  838. ,
  839. std::optional<uint64_t> // lastAuthorizedTime
  840. ,
  841. std::optional<uint64_t> // lastDeauthorizedTime
  842. ,
  843. std::optional<int> // remoteTraceLevel
  844. ,
  845. std::optional<std::string> // remoteTraceTarget
  846. ,
  847. std::optional<std::string> // tags
  848. ,
  849. std::optional<int> // vMajor
  850. ,
  851. std::optional<int> // vMinor
  852. ,
  853. std::optional<int> // vRev
  854. ,
  855. std::optional<int> // vProto
  856. ,
  857. std::optional<bool> // noAutoAssignIps
  858. ,
  859. std::optional<uint64_t> // revision
  860. ,
  861. std::optional<bool> // ssoExempt
  862. ,
  863. std::optional<uint64_t> // authenticationExpiryTime
  864. ,
  865. std::string // assignedAddresses
  866. >
  867. row;
  868. uint64_t count = 0;
  869. auto tmp = std::chrono::high_resolution_clock::now();
  870. uint64_t total = 0;
  871. while (stream >> row) {
  872. auto start = std::chrono::high_resolution_clock::now();
  873. json empty;
  874. json config;
  875. initMember(config);
  876. memberId = std::get<0>(row);
  877. networkId = std::get<1>(row);
  878. std::optional<bool> activeBridge = std::get<2>(row);
  879. std::optional<bool> authorized = std::get<3>(row);
  880. std::optional<std::string> capabilities = std::get<4>(row);
  881. std::optional<uint64_t> creationTime = std::get<5>(row);
  882. std::optional<std::string> identity = std::get<6>(row);
  883. std::optional<uint64_t> lastAuthorizedTime = std::get<7>(row);
  884. std::optional<uint64_t> lastDeauthorizedTime = std::get<8>(row);
  885. std::optional<int> remoteTraceLevel = std::get<9>(row);
  886. std::optional<std::string> remoteTraceTarget = std::get<10>(row);
  887. std::optional<std::string> tags = std::get<11>(row);
  888. std::optional<int> vMajor = std::get<12>(row);
  889. std::optional<int> vMinor = std::get<13>(row);
  890. std::optional<int> vRev = std::get<14>(row);
  891. std::optional<int> vProto = std::get<15>(row);
  892. std::optional<bool> noAutoAssignIps = std::get<16>(row);
  893. std::optional<uint64_t> revision = std::get<17>(row);
  894. std::optional<bool> ssoExempt = std::get<18>(row);
  895. std::optional<uint64_t> authenticationExpiryTime = std::get<19>(row);
  896. std::string assignedAddresses = std::get<20>(row);
  897. networkMembers.insert(std::pair<std::string, std::string>(setKeyBase + networkId, memberId));
  898. config["id"] = memberId;
  899. config["address"] = memberId;
  900. config["nwid"] = networkId;
  901. config["activeBridge"] = activeBridge.value_or(false);
  902. config["authorized"] = authorized.value_or(false);
  903. config["capabilities"] = json::parse(capabilities.value_or("[]"));
  904. config["creationTime"] = creationTime.value_or(0);
  905. config["identity"] = identity.value_or("");
  906. config["lastAuthorizedTime"] = lastAuthorizedTime.value_or(0);
  907. config["lastDeauthorizedTime"] = lastDeauthorizedTime.value_or(0);
  908. config["remoteTraceLevel"] = remoteTraceLevel.value_or(0);
  909. config["remoteTraceTarget"] = remoteTraceTarget.value_or("");
  910. config["tags"] = json::parse(tags.value_or("[]"));
  911. config["vMajor"] = vMajor.value_or(-1);
  912. config["vMinor"] = vMinor.value_or(-1);
  913. config["vRev"] = vRev.value_or(-1);
  914. config["vProto"] = vProto.value_or(-1);
  915. config["noAutoAssignIps"] = noAutoAssignIps.value_or(false);
  916. config["revision"] = revision.value_or(0);
  917. config["ssoExempt"] = ssoExempt.value_or(false);
  918. config["authenticationExpiryTime"] = authenticationExpiryTime.value_or(0);
  919. config["objtype"] = "member";
  920. config["ipAssignments"] = json::array();
  921. if (assignedAddresses != "{}") {
  922. std::string tmp = assignedAddresses.substr(1, assignedAddresses.size() - 2);
  923. std::vector<std::string> addrs = split(tmp, ',');
  924. for (auto it = addrs.begin(); it != addrs.end(); ++it) {
  925. config["ipAssignments"].push_back(*it);
  926. }
  927. }
  928. Metrics::member_count++;
  929. _memberChanged(empty, config, false);
  930. memberId = "";
  931. networkId = "";
  932. auto end = std::chrono::high_resolution_clock::now();
  933. auto dur = std::chrono::duration_cast<std::chrono::microseconds>(end - start);
  934. total += dur.count();
  935. ++count;
  936. if (count > 0 && count % 10000 == 0) {
  937. fprintf(stderr, "Averaging %llu us per member\n", (total / count));
  938. }
  939. }
  940. if (count > 0) {
  941. fprintf(stderr, "Took %llu us per member to load\n", (total / count));
  942. }
  943. stream.complete();
  944. w.commit();
  945. _pool->unborrow(c2);
  946. _pool->unborrow(c);
  947. fprintf(stderr, "done.\n");
  948. if (! networkMembers.empty()) {
  949. if (_redisMemberStatus) {
  950. fprintf(stderr, "Load member data into redis...\n");
  951. if (_rc->clusterMode) {
  952. auto tx = _cluster->transaction(_myAddressStr, true, false);
  953. uint64_t count = 0;
  954. for (auto it : networkMembers) {
  955. tx.sadd(it.first, it.second);
  956. if (++count % 30000 == 0) {
  957. tx.exec();
  958. tx = _cluster->transaction(_myAddressStr, true, false);
  959. }
  960. }
  961. tx.exec();
  962. }
  963. else {
  964. auto tx = _redis->transaction(true, false);
  965. uint64_t count = 0;
  966. for (auto it : networkMembers) {
  967. tx.sadd(it.first, it.second);
  968. if (++count % 30000 == 0) {
  969. tx.exec();
  970. tx = _redis->transaction(true, false);
  971. }
  972. }
  973. tx.exec();
  974. }
  975. fprintf(stderr, "done.\n");
  976. }
  977. }
  978. fprintf(stderr, "Done loading members...\n");
  979. if (++this->_ready == 2) {
  980. if (_waitNoticePrinted) {
  981. fprintf(stderr, "[%s] NOTICE: %.10llx controller PostgreSQL data download complete." ZT_EOL_S, _timestr(), (unsigned long long)_myAddress.toInt());
  982. }
  983. _readyLock.unlock();
  984. }
  985. }
  986. catch (sw::redis::Error& e) {
  987. span->SetStatus(opentelemetry::trace::StatusCode::kError, e.what());
  988. fprintf(stderr, "ERROR: Error initializing members (redis): %s\n", e.what());
  989. exit(-1);
  990. }
  991. catch (std::exception& e) {
  992. span->SetStatus(opentelemetry::trace::StatusCode::kError, e.what());
  993. fprintf(stderr, "ERROR: Error initializing member: %s-%s %s\n", networkId.c_str(), memberId.c_str(), e.what());
  994. exit(-1);
  995. }
  996. }
  997. void CV1::heartbeat()
  998. {
  999. char publicId[1024];
  1000. char hostnameTmp[1024];
  1001. _myId.toString(false, publicId);
  1002. if (gethostname(hostnameTmp, sizeof(hostnameTmp)) != 0) {
  1003. hostnameTmp[0] = (char)0;
  1004. }
  1005. else {
  1006. for (int i = 0; i < (int)sizeof(hostnameTmp); ++i) {
  1007. if ((hostnameTmp[i] == '.') || (hostnameTmp[i] == 0)) {
  1008. hostnameTmp[i] = (char)0;
  1009. break;
  1010. }
  1011. }
  1012. }
  1013. const char* controllerId = _myAddressStr.c_str();
  1014. const char* publicIdentity = publicId;
  1015. const char* hostname = hostnameTmp;
  1016. while (_run == 1) {
  1017. auto provider = opentelemetry::trace::Provider::GetTracerProvider();
  1018. auto tracer = provider->GetTracer("cv1");
  1019. auto span = tracer->StartSpan("cv1::heartbeat");
  1020. auto scope = tracer->WithActiveSpan(span);
  1021. // fprintf(stderr, "%s: heartbeat\n", controllerId);
  1022. auto c = _pool->borrow();
  1023. int64_t ts = OSUtils::now();
  1024. if (c->c) {
  1025. std::string major = std::to_string(ZEROTIER_ONE_VERSION_MAJOR);
  1026. std::string minor = std::to_string(ZEROTIER_ONE_VERSION_MINOR);
  1027. std::string rev = std::to_string(ZEROTIER_ONE_VERSION_REVISION);
  1028. std::string build = std::to_string(ZEROTIER_ONE_VERSION_BUILD);
  1029. std::string now = std::to_string(ts);
  1030. std::string host_port = std::to_string(_listenPort);
  1031. std::string use_redis = (_rc != NULL) ? "true" : "false";
  1032. std::string redis_mem_status = (_redisMemberStatus) ? "true" : "false";
  1033. try {
  1034. pqxx::work w { *c->c };
  1035. pqxx::result res = w.exec0(
  1036. "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) "
  1037. "VALUES ("
  1038. + w.quote(controllerId) + ", " + w.quote(hostname) + ", TO_TIMESTAMP(" + now + "::double precision/1000), " + w.quote(publicIdentity) + ", " + major + ", " + minor + ", " + rev + ", " + build + ", " + host_port + ", "
  1039. + use_redis + ", " + redis_mem_status
  1040. + ") "
  1041. "ON CONFLICT (id) DO UPDATE SET cluster_host = EXCLUDED.cluster_host, last_alive = EXCLUDED.last_alive, "
  1042. "public_identity = EXCLUDED.public_identity, v_major = EXCLUDED.v_major, v_minor = EXCLUDED.v_minor, "
  1043. "v_rev = EXCLUDED.v_rev, v_build = EXCLUDED.v_rev, host_port = EXCLUDED.host_port, "
  1044. "use_redis = EXCLUDED.use_redis, redis_member_status = EXCLUDED.redis_member_status");
  1045. w.commit();
  1046. }
  1047. catch (std::exception& e) {
  1048. fprintf(stderr, "%s: Heartbeat update failed: %s\n", controllerId, e.what());
  1049. span->End();
  1050. std::this_thread::sleep_for(std::chrono::milliseconds(1000));
  1051. continue;
  1052. }
  1053. }
  1054. _pool->unborrow(c);
  1055. try {
  1056. if (_redisMemberStatus) {
  1057. if (_rc->clusterMode) {
  1058. _cluster->zadd("controllers", "controllerId", ts);
  1059. }
  1060. else {
  1061. _redis->zadd("controllers", "controllerId", ts);
  1062. }
  1063. }
  1064. }
  1065. catch (sw::redis::Error& e) {
  1066. fprintf(stderr, "ERROR: Redis error in heartbeat thread: %s\n", e.what());
  1067. }
  1068. span->End();
  1069. std::this_thread::sleep_for(std::chrono::milliseconds(1000));
  1070. }
  1071. fprintf(stderr, "Exited heartbeat thread\n");
  1072. }
  1073. void CV1::membersDbWatcher()
  1074. {
  1075. if (_rc) {
  1076. _membersWatcher_Redis();
  1077. }
  1078. else {
  1079. _membersWatcher_Postgres();
  1080. }
  1081. if (_run == 1) {
  1082. fprintf(stderr, "ERROR: %s membersDbWatcher should still be running! Exiting Controller.\n", _myAddressStr.c_str());
  1083. exit(9);
  1084. }
  1085. fprintf(stderr, "Exited membersDbWatcher\n");
  1086. }
  1087. void CV1::_membersWatcher_Postgres()
  1088. {
  1089. auto c = _pool->borrow();
  1090. std::string stream = "member_" + _myAddressStr;
  1091. fprintf(stderr, "Listening to member stream: %s\n", stream.c_str());
  1092. MemberNotificationReceiver<CV1> m(this, *c->c, stream);
  1093. while (_run == 1) {
  1094. c->c->await_notification(5, 0);
  1095. }
  1096. _pool->unborrow(c);
  1097. }
  1098. void CV1::_membersWatcher_Redis()
  1099. {
  1100. char buf[11] = { 0 };
  1101. std::string key = "member-stream:{" + std::string(_myAddress.toString(buf)) + "}";
  1102. std::string lastID = "0";
  1103. fprintf(stderr, "Listening to member stream: %s\n", key.c_str());
  1104. while (_run == 1) {
  1105. auto provider = opentelemetry::trace::Provider::GetTracerProvider();
  1106. auto tracer = provider->GetTracer("cv1");
  1107. auto span = tracer->StartSpan("cv1::_membersWatcher_Redis");
  1108. auto scope = tracer->WithActiveSpan(span);
  1109. try {
  1110. json tmp;
  1111. std::unordered_map<std::string, ItemStream> result;
  1112. if (_rc->clusterMode) {
  1113. _cluster->xread(key, lastID, std::chrono::seconds(1), 0, std::inserter(result, result.end()));
  1114. }
  1115. else {
  1116. _redis->xread(key, lastID, std::chrono::seconds(1), 0, std::inserter(result, result.end()));
  1117. }
  1118. if (! result.empty()) {
  1119. for (auto element : result) {
  1120. #ifdef REDIS_TRACE
  1121. fprintf(stdout, "Received notification from: %s\n", element.first.c_str());
  1122. #endif
  1123. for (auto rec : element.second) {
  1124. std::string id = rec.first;
  1125. auto attrs = rec.second;
  1126. #ifdef REDIS_TRACE
  1127. fprintf(stdout, "Record ID: %s\n", id.c_str());
  1128. fprintf(stdout, "attrs len: %lu\n", attrs.size());
  1129. #endif
  1130. for (auto a : attrs) {
  1131. #ifdef REDIS_TRACE
  1132. fprintf(stdout, "key: %s\nvalue: %s\n", a.first.c_str(), a.second.c_str());
  1133. #endif
  1134. try {
  1135. tmp = json::parse(a.second);
  1136. json& ov = tmp["old_val"];
  1137. json& nv = tmp["new_val"];
  1138. json oldConfig, newConfig;
  1139. if (ov.is_object())
  1140. oldConfig = ov;
  1141. if (nv.is_object())
  1142. newConfig = nv;
  1143. if (oldConfig.is_object() || newConfig.is_object()) {
  1144. _memberChanged(oldConfig, newConfig, (this->_ready >= 2));
  1145. }
  1146. }
  1147. catch (...) {
  1148. fprintf(stderr, "json parse error in _membersWatcher_Redis: %s\n", a.second.c_str());
  1149. }
  1150. }
  1151. if (_rc->clusterMode) {
  1152. _cluster->xdel(key, id);
  1153. }
  1154. else {
  1155. _redis->xdel(key, id);
  1156. }
  1157. lastID = id;
  1158. Metrics::redis_mem_notification++;
  1159. }
  1160. }
  1161. }
  1162. }
  1163. catch (sw::redis::Error& e) {
  1164. span->SetStatus(opentelemetry::trace::StatusCode::kError, e.what());
  1165. fprintf(stderr, "Error in Redis members watcher: %s\n", e.what());
  1166. }
  1167. }
  1168. fprintf(stderr, "membersWatcher ended\n");
  1169. }
  1170. void CV1::networksDbWatcher()
  1171. {
  1172. if (_rc) {
  1173. _networksWatcher_Redis();
  1174. }
  1175. else {
  1176. _networksWatcher_Postgres();
  1177. }
  1178. if (_run == 1) {
  1179. fprintf(stderr, "ERROR: %s networksDbWatcher should still be running! Exiting Controller.\n", _myAddressStr.c_str());
  1180. exit(8);
  1181. }
  1182. fprintf(stderr, "Exited networksDbWatcher\n");
  1183. }
  1184. void CV1::_networksWatcher_Postgres()
  1185. {
  1186. std::string stream = "network_" + _myAddressStr;
  1187. fprintf(stderr, "Listening to member stream: %s\n", stream.c_str());
  1188. auto c = _pool->borrow();
  1189. NetworkNotificationReceiver<CV1> n(this, *c->c, stream);
  1190. while (_run == 1) {
  1191. auto provider = opentelemetry::trace::Provider::GetTracerProvider();
  1192. auto tracer = provider->GetTracer("cv1");
  1193. auto span = tracer->StartSpan("cv1::_networksWatcher_Postgres");
  1194. auto scope = tracer->WithActiveSpan(span);
  1195. c->c->await_notification(5, 0);
  1196. }
  1197. }
  1198. void CV1::_networksWatcher_Redis()
  1199. {
  1200. char buf[11] = { 0 };
  1201. std::string key = "network-stream:{" + std::string(_myAddress.toString(buf)) + "}";
  1202. std::string lastID = "0";
  1203. while (_run == 1) {
  1204. auto provider = opentelemetry::trace::Provider::GetTracerProvider();
  1205. auto tracer = provider->GetTracer("cv1");
  1206. auto span = tracer->StartSpan("cv1::_networksWatcher_Redis");
  1207. auto scope = tracer->WithActiveSpan(span);
  1208. try {
  1209. json tmp;
  1210. std::unordered_map<std::string, ItemStream> result;
  1211. if (_rc->clusterMode) {
  1212. _cluster->xread(key, lastID, std::chrono::seconds(1), 0, std::inserter(result, result.end()));
  1213. }
  1214. else {
  1215. _redis->xread(key, lastID, std::chrono::seconds(1), 0, std::inserter(result, result.end()));
  1216. }
  1217. if (! result.empty()) {
  1218. for (auto element : result) {
  1219. #ifdef REDIS_TRACE
  1220. fprintf(stdout, "Received notification from: %s\n", element.first.c_str());
  1221. #endif
  1222. for (auto rec : element.second) {
  1223. std::string id = rec.first;
  1224. auto attrs = rec.second;
  1225. #ifdef REDIS_TRACE
  1226. fprintf(stdout, "Record ID: %s\n", id.c_str());
  1227. fprintf(stdout, "attrs len: %lu\n", attrs.size());
  1228. #endif
  1229. for (auto a : attrs) {
  1230. #ifdef REDIS_TRACE
  1231. fprintf(stdout, "key: %s\nvalue: %s\n", a.first.c_str(), a.second.c_str());
  1232. #endif
  1233. try {
  1234. tmp = json::parse(a.second);
  1235. json& ov = tmp["old_val"];
  1236. json& nv = tmp["new_val"];
  1237. json oldConfig, newConfig;
  1238. if (ov.is_object())
  1239. oldConfig = ov;
  1240. if (nv.is_object())
  1241. newConfig = nv;
  1242. if (oldConfig.is_object() || newConfig.is_object()) {
  1243. _networkChanged(oldConfig, newConfig, (this->_ready >= 2));
  1244. }
  1245. }
  1246. catch (std::exception& e) {
  1247. fprintf(stderr, "json parse error in networkWatcher_Redis: what: %s json: %s\n", e.what(), a.second.c_str());
  1248. }
  1249. }
  1250. if (_rc->clusterMode) {
  1251. _cluster->xdel(key, id);
  1252. }
  1253. else {
  1254. _redis->xdel(key, id);
  1255. }
  1256. lastID = id;
  1257. }
  1258. Metrics::redis_net_notification++;
  1259. }
  1260. }
  1261. }
  1262. catch (sw::redis::Error& e) {
  1263. span->SetStatus(opentelemetry::trace::StatusCode::kError, e.what());
  1264. fprintf(stderr, "Error in Redis networks watcher: %s\n", e.what());
  1265. }
  1266. }
  1267. fprintf(stderr, "networksWatcher ended\n");
  1268. }
  1269. void CV1::commitThread()
  1270. {
  1271. fprintf(stderr, "%s: commitThread start\n", _myAddressStr.c_str());
  1272. std::pair<nlohmann::json, bool> qitem;
  1273. while (_commitQueue.get(qitem) & (_run == 1)) {
  1274. auto provider = opentelemetry::trace::Provider::GetTracerProvider();
  1275. auto tracer = provider->GetTracer("cv1");
  1276. auto span = tracer->StartSpan("cv1::commitThread");
  1277. auto scope = tracer->WithActiveSpan(span);
  1278. // fprintf(stderr, "commitThread tick\n");
  1279. if (! qitem.first.is_object()) {
  1280. fprintf(stderr, "not an object\n");
  1281. continue;
  1282. }
  1283. std::shared_ptr<PostgresConnection> c;
  1284. try {
  1285. c = _pool->borrow();
  1286. }
  1287. catch (std::exception& e) {
  1288. fprintf(stderr, "ERROR: %s\n", e.what());
  1289. continue;
  1290. }
  1291. if (! c) {
  1292. fprintf(stderr, "Error getting database connection\n");
  1293. continue;
  1294. }
  1295. Metrics::pgsql_commit_ticks++;
  1296. try {
  1297. nlohmann::json& config = (qitem.first);
  1298. const std::string objtype = config["objtype"];
  1299. if (objtype == "member") {
  1300. auto mspan = tracer->StartSpan("cv1::commitThread::member");
  1301. auto mscope = tracer->WithActiveSpan(mspan);
  1302. // fprintf(stderr, "%s: commitThread: member\n", _myAddressStr.c_str());
  1303. std::string memberId;
  1304. std::string networkId;
  1305. try {
  1306. pqxx::work w(*c->c);
  1307. memberId = config["id"];
  1308. networkId = config["nwid"];
  1309. std::string target = "NULL";
  1310. if (! config["remoteTraceTarget"].is_null()) {
  1311. target = config["remoteTraceTarget"];
  1312. }
  1313. pqxx::row nwrow = w.exec_params1("SELECT COUNT(id) FROM ztc_network WHERE id = $1", networkId);
  1314. int nwcount = nwrow[0].as<int>();
  1315. if (nwcount != 1) {
  1316. fprintf(stderr, "network %s does not exist. skipping member upsert\n", networkId.c_str());
  1317. w.abort();
  1318. _pool->unborrow(c);
  1319. continue;
  1320. }
  1321. pqxx::row mrow = w.exec_params1("SELECT COUNT(id) FROM ztc_member WHERE id = $1 AND network_id = $2", memberId, networkId);
  1322. int membercount = mrow[0].as<int>();
  1323. bool isNewMember = false;
  1324. if (membercount == 0) {
  1325. // new member
  1326. isNewMember = true;
  1327. pqxx::result res = w.exec_params0(
  1328. "INSERT INTO ztc_member (id, network_id, active_bridge, authorized, capabilities, "
  1329. "identity, last_authorized_time, last_deauthorized_time, no_auto_assign_ips, "
  1330. "remote_trace_level, remote_trace_target, revision, tags, v_major, v_minor, v_rev, v_proto) "
  1331. "VALUES ($1, $2, $3, $4, $5, $6, "
  1332. "TO_TIMESTAMP($7::double precision/1000), TO_TIMESTAMP($8::double precision/1000), "
  1333. "$9, $10, $11, $12, $13, $14, $15, $16, $17)",
  1334. memberId,
  1335. networkId,
  1336. (bool)config["activeBridge"],
  1337. (bool)config["authorized"],
  1338. OSUtils::jsonDump(config["capabilities"], -1),
  1339. OSUtils::jsonString(config["identity"], ""),
  1340. (uint64_t)config["lastAuthorizedTime"],
  1341. (uint64_t)config["lastDeauthorizedTime"],
  1342. (bool)config["noAutoAssignIps"],
  1343. (int)config["remoteTraceLevel"],
  1344. target,
  1345. (uint64_t)config["revision"],
  1346. OSUtils::jsonDump(config["tags"], -1),
  1347. (int)config["vMajor"],
  1348. (int)config["vMinor"],
  1349. (int)config["vRev"],
  1350. (int)config["vProto"]);
  1351. }
  1352. else {
  1353. // existing member
  1354. pqxx::result res = w.exec_params0(
  1355. "UPDATE ztc_member "
  1356. "SET active_bridge = $3, authorized = $4, capabilities = $5, identity = $6, "
  1357. "last_authorized_time = TO_TIMESTAMP($7::double precision/1000), "
  1358. "last_deauthorized_time = TO_TIMESTAMP($8::double precision/1000), "
  1359. "no_auto_assign_ips = $9, remote_trace_level = $10, remote_trace_target= $11, "
  1360. "revision = $12, tags = $13, v_major = $14, v_minor = $15, v_rev = $16, v_proto = $17 "
  1361. "WHERE id = $1 AND network_id = $2",
  1362. memberId,
  1363. networkId,
  1364. (bool)config["activeBridge"],
  1365. (bool)config["authorized"],
  1366. OSUtils::jsonDump(config["capabilities"], -1),
  1367. OSUtils::jsonString(config["identity"], ""),
  1368. (uint64_t)config["lastAuthorizedTime"],
  1369. (uint64_t)config["lastDeauthorizedTime"],
  1370. (bool)config["noAutoAssignIps"],
  1371. (int)config["remoteTraceLevel"],
  1372. target,
  1373. (uint64_t)config["revision"],
  1374. OSUtils::jsonDump(config["tags"], -1),
  1375. (int)config["vMajor"],
  1376. (int)config["vMinor"],
  1377. (int)config["vRev"],
  1378. (int)config["vProto"]);
  1379. }
  1380. if (! isNewMember) {
  1381. pqxx::result res = w.exec_params0("DELETE FROM ztc_member_ip_assignment WHERE member_id = $1 AND network_id = $2", memberId, networkId);
  1382. }
  1383. std::vector<std::string> assignments;
  1384. bool ipAssignError = false;
  1385. for (auto i = config["ipAssignments"].begin(); i != config["ipAssignments"].end(); ++i) {
  1386. std::string addr = *i;
  1387. if (std::find(assignments.begin(), assignments.end(), addr) != assignments.end()) {
  1388. continue;
  1389. }
  1390. 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);
  1391. assignments.push_back(addr);
  1392. }
  1393. if (ipAssignError) {
  1394. fprintf(stderr, "%s: ipAssignError\n", _myAddressStr.c_str());
  1395. w.abort();
  1396. _pool->unborrow(c);
  1397. c.reset();
  1398. continue;
  1399. }
  1400. w.commit();
  1401. if (_smee != NULL && isNewMember) {
  1402. pqxx::row row = w.exec_params1(
  1403. "SELECT "
  1404. " count(h.hook_id) "
  1405. "FROM "
  1406. " ztc_hook h "
  1407. " INNER JOIN ztc_org o ON o.org_id = h.org_id "
  1408. " INNER JOIN ztc_network n ON n.owner_id = o.owner_id "
  1409. " WHERE "
  1410. "n.id = $1 ",
  1411. networkId);
  1412. int64_t hookCount = row[0].as<int64_t>();
  1413. if (hookCount > 0) {
  1414. notifyNewMember(networkId, memberId);
  1415. }
  1416. }
  1417. const uint64_t nwidInt = OSUtils::jsonIntHex(config["nwid"], 0ULL);
  1418. const uint64_t memberidInt = OSUtils::jsonIntHex(config["id"], 0ULL);
  1419. if (nwidInt && memberidInt) {
  1420. nlohmann::json nwOrig;
  1421. nlohmann::json memOrig;
  1422. nlohmann::json memNew(config);
  1423. get(nwidInt, nwOrig, memberidInt, memOrig);
  1424. _memberChanged(memOrig, memNew, qitem.second);
  1425. }
  1426. else {
  1427. 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);
  1428. }
  1429. }
  1430. catch (std::exception& e) {
  1431. fprintf(stderr, "%s ERROR: Error updating member %s-%s: %s\n", _myAddressStr.c_str(), networkId.c_str(), memberId.c_str(), e.what());
  1432. mspan->SetStatus(opentelemetry::trace::StatusCode::kError, e.what());
  1433. }
  1434. }
  1435. else if (objtype == "network") {
  1436. auto nspan = tracer->StartSpan("cv1::commitThread::network");
  1437. auto nscope = tracer->WithActiveSpan(nspan);
  1438. try {
  1439. // fprintf(stderr, "%s: commitThread: network\n", _myAddressStr.c_str());
  1440. pqxx::work w(*c->c);
  1441. std::string id = config["id"];
  1442. std::string remoteTraceTarget = "";
  1443. if (! config["remoteTraceTarget"].is_null()) {
  1444. remoteTraceTarget = config["remoteTraceTarget"];
  1445. }
  1446. std::string rulesSource = "";
  1447. if (config["rulesSource"].is_string()) {
  1448. rulesSource = config["rulesSource"];
  1449. }
  1450. // This ugly query exists because when we want to mirror networks to/from
  1451. // another data store (e.g. FileDB or LFDB) it is possible to get a network
  1452. // that doesn't exist in Central's database. This does an upsert and sets
  1453. // the owner_id to the "first" global admin in the user DB if the record
  1454. // did not previously exist. If the record already exists owner_id is left
  1455. // unchanged, so owner_id should be left out of the update clause.
  1456. pqxx::result res = w.exec_params0(
  1457. "INSERT INTO ztc_network (id, creation_time, owner_id, controller_id, capabilities, enable_broadcast, "
  1458. "last_modified, mtu, multicast_limit, name, private, "
  1459. "remote_trace_level, remote_trace_target, rules, rules_source, "
  1460. "tags, v4_assign_mode, v6_assign_mode, sso_enabled) VALUES ("
  1461. "$1, TO_TIMESTAMP($5::double precision/1000), "
  1462. "(SELECT user_id AS owner_id FROM ztc_global_permissions WHERE authorize = true AND del = true AND modify = true AND read = true LIMIT 1),"
  1463. "$2, $3, $4, TO_TIMESTAMP($5::double precision/1000), "
  1464. "$6, $7, $8, $9, $10, $11, $12, $13, $14, $15, $16, $17) "
  1465. "ON CONFLICT (id) DO UPDATE set controller_id = EXCLUDED.controller_id, "
  1466. "capabilities = EXCLUDED.capabilities, enable_broadcast = EXCLUDED.enable_broadcast, "
  1467. "last_modified = EXCLUDED.last_modified, mtu = EXCLUDED.mtu, "
  1468. "multicast_limit = EXCLUDED.multicast_limit, name = EXCLUDED.name, "
  1469. "private = EXCLUDED.private, remote_trace_level = EXCLUDED.remote_trace_level, "
  1470. "remote_trace_target = EXCLUDED.remote_trace_target, rules = EXCLUDED.rules, "
  1471. "rules_source = EXCLUDED.rules_source, tags = EXCLUDED.tags, "
  1472. "v4_assign_mode = EXCLUDED.v4_assign_mode, v6_assign_mode = EXCLUDED.v6_assign_mode, "
  1473. "sso_enabled = EXCLUDED.sso_enabled",
  1474. id,
  1475. _myAddressStr,
  1476. OSUtils::jsonDump(config["capabilities"], -1),
  1477. (bool)config["enableBroadcast"],
  1478. OSUtils::now(),
  1479. (int)config["mtu"],
  1480. (int)config["multicastLimit"],
  1481. OSUtils::jsonString(config["name"], ""),
  1482. (bool)config["private"],
  1483. (int)config["remoteTraceLevel"],
  1484. remoteTraceTarget,
  1485. OSUtils::jsonDump(config["rules"], -1),
  1486. rulesSource,
  1487. OSUtils::jsonDump(config["tags"], -1),
  1488. OSUtils::jsonDump(config["v4AssignMode"], -1),
  1489. OSUtils::jsonDump(config["v6AssignMode"], -1),
  1490. OSUtils::jsonBool(config["ssoEnabled"], false));
  1491. res = w.exec_params0("DELETE FROM ztc_network_assignment_pool WHERE network_id = $1", 0);
  1492. auto pool = config["ipAssignmentPools"];
  1493. bool err = false;
  1494. for (auto i = pool.begin(); i != pool.end(); ++i) {
  1495. std::string start = (*i)["ipRangeStart"];
  1496. std::string end = (*i)["ipRangeEnd"];
  1497. res = w.exec_params0(
  1498. "INSERT INTO ztc_network_assignment_pool (network_id, ip_range_start, ip_range_end) "
  1499. "VALUES ($1, $2, $3)",
  1500. id,
  1501. start,
  1502. end);
  1503. }
  1504. res = w.exec_params0("DELETE FROM ztc_network_route WHERE network_id = $1", id);
  1505. auto routes = config["routes"];
  1506. err = false;
  1507. for (auto i = routes.begin(); i != routes.end(); ++i) {
  1508. std::string t = (*i)["target"];
  1509. std::vector<std::string> target;
  1510. std::istringstream f(t);
  1511. std::string s;
  1512. while (std::getline(f, s, '/')) {
  1513. target.push_back(s);
  1514. }
  1515. if (target.empty() || target.size() != 2) {
  1516. continue;
  1517. }
  1518. std::string targetAddr = target[0];
  1519. std::string targetBits = target[1];
  1520. std::string via = "NULL";
  1521. if (! (*i)["via"].is_null()) {
  1522. via = (*i)["via"];
  1523. }
  1524. 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()));
  1525. }
  1526. if (err) {
  1527. fprintf(stderr, "%s: route add error\n", _myAddressStr.c_str());
  1528. w.abort();
  1529. _pool->unborrow(c);
  1530. continue;
  1531. }
  1532. auto dns = config["dns"];
  1533. std::string domain = dns["domain"];
  1534. std::stringstream servers;
  1535. servers << "{";
  1536. for (auto j = dns["servers"].begin(); j < dns["servers"].end(); ++j) {
  1537. servers << *j;
  1538. if ((j + 1) != dns["servers"].end()) {
  1539. servers << ",";
  1540. }
  1541. }
  1542. servers << "}";
  1543. std::string s = servers.str();
  1544. 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);
  1545. w.commit();
  1546. const uint64_t nwidInt = OSUtils::jsonIntHex(config["nwid"], 0ULL);
  1547. if (nwidInt) {
  1548. nlohmann::json nwOrig;
  1549. nlohmann::json nwNew(config);
  1550. get(nwidInt, nwOrig);
  1551. _networkChanged(nwOrig, nwNew, qitem.second);
  1552. }
  1553. else {
  1554. fprintf(stderr, "%s: Can't notify network changed: %llu\n", _myAddressStr.c_str(), (unsigned long long)nwidInt);
  1555. }
  1556. }
  1557. catch (std::exception& e) {
  1558. nspan->SetStatus(opentelemetry::trace::StatusCode::kError, e.what());
  1559. fprintf(stderr, "%s ERROR: Error updating network: %s\n", _myAddressStr.c_str(), e.what());
  1560. }
  1561. if (_redisMemberStatus) {
  1562. try {
  1563. std::string id = config["id"];
  1564. std::string controllerId = _myAddressStr.c_str();
  1565. std::string key = "networks:{" + controllerId + "}";
  1566. if (_rc->clusterMode) {
  1567. _cluster->sadd(key, id);
  1568. }
  1569. else {
  1570. _redis->sadd(key, id);
  1571. }
  1572. }
  1573. catch (sw::redis::Error& e) {
  1574. nspan->SetStatus(opentelemetry::trace::StatusCode::kError, e.what());
  1575. fprintf(stderr, "ERROR: Error adding network to Redis: %s\n", e.what());
  1576. }
  1577. }
  1578. }
  1579. else if (objtype == "_delete_network") {
  1580. auto dspan = tracer->StartSpan("cv1::commitThread::_delete_network");
  1581. auto dscope = tracer->WithActiveSpan(dspan);
  1582. // fprintf(stderr, "%s: commitThread: delete network\n", _myAddressStr.c_str());
  1583. try {
  1584. pqxx::work w(*c->c);
  1585. std::string networkId = config["nwid"];
  1586. pqxx::result res = w.exec_params0("UPDATE ztc_network SET deleted = true WHERE id = $1", networkId);
  1587. w.commit();
  1588. }
  1589. catch (std::exception& e) {
  1590. dspan->SetStatus(opentelemetry::trace::StatusCode::kError, e.what());
  1591. fprintf(stderr, "%s ERROR: Error deleting network: %s\n", _myAddressStr.c_str(), e.what());
  1592. }
  1593. if (_redisMemberStatus) {
  1594. try {
  1595. std::string id = config["id"];
  1596. std::string controllerId = _myAddressStr.c_str();
  1597. std::string key = "networks:{" + controllerId + "}";
  1598. if (_rc->clusterMode) {
  1599. _cluster->srem(key, id);
  1600. _cluster->del("network-nodes-online:{" + controllerId + "}:" + id);
  1601. }
  1602. else {
  1603. _redis->srem(key, id);
  1604. _redis->del("network-nodes-online:{" + controllerId + "}:" + id);
  1605. }
  1606. }
  1607. catch (sw::redis::Error& e) {
  1608. dspan->SetStatus(opentelemetry::trace::StatusCode::kError, e.what());
  1609. fprintf(stderr, "ERROR: Error adding network to Redis: %s\n", e.what());
  1610. }
  1611. }
  1612. }
  1613. else if (objtype == "_delete_member") {
  1614. auto mspan = tracer->StartSpan("cv1::commitThread::_delete_member");
  1615. auto mscope = tracer->WithActiveSpan(mspan);
  1616. // fprintf(stderr, "%s commitThread: delete member\n", _myAddressStr.c_str());
  1617. try {
  1618. pqxx::work w(*c->c);
  1619. std::string memberId = config["id"];
  1620. std::string networkId = config["nwid"];
  1621. pqxx::result res = w.exec_params0("UPDATE ztc_member SET hidden = true, deleted = true WHERE id = $1 AND network_id = $2", memberId, networkId);
  1622. w.commit();
  1623. }
  1624. catch (std::exception& e) {
  1625. mspan->SetStatus(opentelemetry::trace::StatusCode::kError, e.what());
  1626. fprintf(stderr, "%s ERROR: Error deleting member: %s\n", _myAddressStr.c_str(), e.what());
  1627. }
  1628. if (_redisMemberStatus) {
  1629. try {
  1630. std::string memberId = config["id"];
  1631. std::string networkId = config["nwid"];
  1632. std::string controllerId = _myAddressStr.c_str();
  1633. std::string key = "network-nodes-all:{" + controllerId + "}:" + networkId;
  1634. if (_rc->clusterMode) {
  1635. _cluster->srem(key, memberId);
  1636. _cluster->del("member:{" + controllerId + "}:" + networkId + ":" + memberId);
  1637. }
  1638. else {
  1639. _redis->srem(key, memberId);
  1640. _redis->del("member:{" + controllerId + "}:" + networkId + ":" + memberId);
  1641. }
  1642. }
  1643. catch (sw::redis::Error& e) {
  1644. mspan->SetStatus(opentelemetry::trace::StatusCode::kError, e.what());
  1645. fprintf(stderr, "ERROR: Error deleting member from Redis: %s\n", e.what());
  1646. }
  1647. }
  1648. }
  1649. else {
  1650. fprintf(stderr, "%s ERROR: unknown objtype\n", _myAddressStr.c_str());
  1651. }
  1652. }
  1653. catch (std::exception& e) {
  1654. span->SetStatus(opentelemetry::trace::StatusCode::kError, e.what());
  1655. fprintf(stderr, "%s ERROR: Error getting objtype: %s\n", _myAddressStr.c_str(), e.what());
  1656. }
  1657. _pool->unborrow(c);
  1658. c.reset();
  1659. }
  1660. fprintf(stderr, "%s commitThread finished\n", _myAddressStr.c_str());
  1661. }
  1662. void CV1::notifyNewMember(const std::string& networkID, const std::string& memberID)
  1663. {
  1664. auto provider = opentelemetry::trace::Provider::GetTracerProvider();
  1665. auto tracer = provider->GetTracer("cv1");
  1666. auto span = tracer->StartSpan("cv1::notifyNewMember");
  1667. auto scope = tracer->WithActiveSpan(span);
  1668. smeeclient::smee_client_notify_network_joined(_smee, networkID.c_str(), memberID.c_str());
  1669. }
  1670. void CV1::onlineNotificationThread()
  1671. {
  1672. waitForReady();
  1673. if (_redisMemberStatus) {
  1674. onlineNotification_Redis();
  1675. }
  1676. else {
  1677. onlineNotification_Postgres();
  1678. }
  1679. }
  1680. /**
  1681. * ONLY UNCOMMENT FOR TEMPORARY DB MAINTENANCE
  1682. *
  1683. * This define temporarily turns off writing to the member status table
  1684. * so it can be reindexed when the indexes get too large.
  1685. */
  1686. // #define DISABLE_MEMBER_STATUS 1
  1687. void CV1::onlineNotification_Postgres()
  1688. {
  1689. _connected = 1;
  1690. nlohmann::json jtmp1, jtmp2;
  1691. while (_run == 1) {
  1692. auto provider = opentelemetry::trace::Provider::GetTracerProvider();
  1693. auto tracer = provider->GetTracer("cv1");
  1694. auto span = tracer->StartSpan("cv1::onlineNotification_Postgres");
  1695. auto scope = tracer->WithActiveSpan(span);
  1696. auto c = _pool->borrow();
  1697. auto c2 = _pool->borrow();
  1698. try {
  1699. fprintf(stderr, "%s onlineNotification_Postgres\n", _myAddressStr.c_str());
  1700. std::unordered_map<std::pair<uint64_t, uint64_t>, NodeOnlineRecord, _PairHasher> lastOnline;
  1701. {
  1702. std::lock_guard<std::mutex> l(_lastOnline_l);
  1703. lastOnline.swap(_lastOnline);
  1704. }
  1705. #ifndef DISABLE_MEMBER_STATUS
  1706. pqxx::work w(*c->c);
  1707. pqxx::work w2(*c2->c);
  1708. fprintf(stderr, "online notification tick\n");
  1709. bool firstRun = true;
  1710. bool memberAdded = false;
  1711. int updateCount = 0;
  1712. pqxx::pipeline pipe(w);
  1713. for (auto i = lastOnline.begin(); i != lastOnline.end(); ++i) {
  1714. updateCount += 1;
  1715. uint64_t nwid_i = i->first.first;
  1716. char nwidTmp[64];
  1717. char memTmp[64];
  1718. char ipTmp[64];
  1719. OSUtils::ztsnprintf(nwidTmp, sizeof(nwidTmp), "%.16llx", nwid_i);
  1720. OSUtils::ztsnprintf(memTmp, sizeof(memTmp), "%.10llx", i->first.second);
  1721. if (! get(nwid_i, jtmp1, i->first.second, jtmp2)) {
  1722. continue; // skip non existent networks/members
  1723. }
  1724. std::string networkId(nwidTmp);
  1725. std::string memberId(memTmp);
  1726. try {
  1727. pqxx::row r = w2.exec_params1("SELECT id, network_id FROM ztc_member WHERE network_id = $1 AND id = $2", networkId, memberId);
  1728. }
  1729. catch (pqxx::unexpected_rows& e) {
  1730. continue;
  1731. }
  1732. int64_t ts = i->second.lastSeen;
  1733. std::string ipAddr = i->second.physicalAddress.toIpString(ipTmp);
  1734. std::string timestamp = std::to_string(ts);
  1735. std::string osArch = i->second.osArch;
  1736. std::stringstream memberUpdate;
  1737. memberUpdate << "INSERT INTO ztc_member_status (network_id, member_id, address, last_updated) VALUES "
  1738. << "('" << networkId << "', '" << memberId << "', ";
  1739. if (ipAddr.empty()) {
  1740. memberUpdate << "NULL, ";
  1741. }
  1742. else {
  1743. memberUpdate << "'" << ipAddr << "', ";
  1744. }
  1745. memberUpdate << "TO_TIMESTAMP(" << timestamp << "::double precision/1000)) "
  1746. << " ON CONFLICT (network_id, member_id) DO UPDATE SET address = EXCLUDED.address, last_updated = EXCLUDED.last_updated";
  1747. pipe.insert(memberUpdate.str());
  1748. Metrics::pgsql_node_checkin++;
  1749. }
  1750. while (! pipe.empty()) {
  1751. pipe.retrieve();
  1752. }
  1753. pipe.complete();
  1754. w.commit();
  1755. fprintf(stderr, "%s: Updated online status of %d members\n", _myAddressStr.c_str(), updateCount);
  1756. #endif
  1757. }
  1758. catch (std::exception& e) {
  1759. fprintf(stderr, "%s: error in onlinenotification thread: %s\n", _myAddressStr.c_str(), e.what());
  1760. }
  1761. _pool->unborrow(c2);
  1762. _pool->unborrow(c);
  1763. ConnectionPoolStats stats = _pool->get_stats();
  1764. 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));
  1765. span->End();
  1766. std::this_thread::sleep_for(std::chrono::seconds(10));
  1767. }
  1768. fprintf(stderr, "%s: Fell out of run loop in onlineNotificationThread\n", _myAddressStr.c_str());
  1769. if (_run == 1) {
  1770. fprintf(stderr, "ERROR: %s onlineNotificationThread should still be running! Exiting Controller.\n", _myAddressStr.c_str());
  1771. exit(6);
  1772. }
  1773. }
  1774. void CV1::onlineNotification_Redis()
  1775. {
  1776. _connected = 1;
  1777. char buf[11] = { 0 };
  1778. std::string controllerId = std::string(_myAddress.toString(buf));
  1779. while (_run == 1) {
  1780. auto provider = opentelemetry::trace::Provider::GetTracerProvider();
  1781. auto tracer = provider->GetTracer("cv1");
  1782. auto span = tracer->StartSpan("cv1::onlineNotification_Redis");
  1783. auto scope = tracer->WithActiveSpan(span);
  1784. fprintf(stderr, "onlineNotification tick\n");
  1785. auto start = std::chrono::high_resolution_clock::now();
  1786. uint64_t count = 0;
  1787. std::unordered_map<std::pair<uint64_t, uint64_t>, NodeOnlineRecord, _PairHasher> lastOnline;
  1788. {
  1789. std::lock_guard<std::mutex> l(_lastOnline_l);
  1790. lastOnline.swap(_lastOnline);
  1791. }
  1792. try {
  1793. if (! lastOnline.empty()) {
  1794. if (_rc->clusterMode) {
  1795. auto tx = _cluster->transaction(controllerId, true, false);
  1796. count = _doRedisUpdate(tx, controllerId, lastOnline);
  1797. }
  1798. else {
  1799. auto tx = _redis->transaction(true, false);
  1800. count = _doRedisUpdate(tx, controllerId, lastOnline);
  1801. }
  1802. }
  1803. }
  1804. catch (sw::redis::Error& e) {
  1805. fprintf(stderr, "Error in online notification thread (redis): %s\n", e.what());
  1806. }
  1807. auto end = std::chrono::high_resolution_clock::now();
  1808. auto dur = std::chrono::duration_cast<std::chrono::milliseconds>(end - start);
  1809. auto total = dur.count();
  1810. fprintf(stderr, "onlineNotification ran in %llu ms\n", total);
  1811. span->End();
  1812. std::this_thread::sleep_for(std::chrono::seconds(5));
  1813. }
  1814. }
  1815. uint64_t CV1::_doRedisUpdate(sw::redis::Transaction& tx, std::string& controllerId, std::unordered_map<std::pair<uint64_t, uint64_t>, NodeOnlineRecord, _PairHasher>& lastOnline)
  1816. {
  1817. auto provider = opentelemetry::trace::Provider::GetTracerProvider();
  1818. auto tracer = provider->GetTracer("cv1");
  1819. auto span = tracer->StartSpan("cv1::_doRedisUpdate");
  1820. auto scope = tracer->WithActiveSpan(span);
  1821. nlohmann::json jtmp1, jtmp2;
  1822. uint64_t count = 0;
  1823. for (auto i = lastOnline.begin(); i != lastOnline.end(); ++i) {
  1824. uint64_t nwid_i = i->first.first;
  1825. uint64_t memberid_i = i->first.second;
  1826. char nwidTmp[64];
  1827. char memTmp[64];
  1828. char ipTmp[64];
  1829. OSUtils::ztsnprintf(nwidTmp, sizeof(nwidTmp), "%.16llx", nwid_i);
  1830. OSUtils::ztsnprintf(memTmp, sizeof(memTmp), "%.10llx", memberid_i);
  1831. if (! get(nwid_i, jtmp1, memberid_i, jtmp2)) {
  1832. continue; // skip non existent members/networks
  1833. }
  1834. std::string networkId(nwidTmp);
  1835. std::string memberId(memTmp);
  1836. int64_t ts = i->second.lastSeen;
  1837. std::string ipAddr = i->second.physicalAddress.toIpString(ipTmp);
  1838. std::string timestamp = std::to_string(ts);
  1839. std::string osArch = i->second.osArch;
  1840. std::unordered_map<std::string, std::string> record = { { "id", memberId }, { "address", ipAddr }, { "last_updated", std::to_string(ts) } };
  1841. tx.zadd("nodes-online:{" + controllerId + "}", memberId, ts)
  1842. .zadd("nodes-online2:{" + controllerId + "}", networkId + "-" + memberId, ts)
  1843. .zadd("network-nodes-online:{" + controllerId + "}:" + networkId, memberId, ts)
  1844. .zadd("active-networks:{" + controllerId + "}", networkId, ts)
  1845. .sadd("network-nodes-all:{" + controllerId + "}:" + networkId, memberId)
  1846. .hmset("member:{" + controllerId + "}:" + networkId + ":" + memberId, record.begin(), record.end());
  1847. ++count;
  1848. Metrics::redis_node_checkin++;
  1849. }
  1850. // expire records from all-nodes and network-nodes member list
  1851. uint64_t expireOld = OSUtils::now() - 300000;
  1852. tx.zremrangebyscore("nodes-online:{" + controllerId + "}", sw::redis::RightBoundedInterval<double>(expireOld, sw::redis::BoundType::LEFT_OPEN));
  1853. tx.zremrangebyscore("nodes-online2:{" + controllerId + "}", sw::redis::RightBoundedInterval<double>(expireOld, sw::redis::BoundType::LEFT_OPEN));
  1854. tx.zremrangebyscore("active-networks:{" + controllerId + "}", sw::redis::RightBoundedInterval<double>(expireOld, sw::redis::BoundType::LEFT_OPEN));
  1855. {
  1856. std::shared_lock<std::shared_mutex> l(_networks_l);
  1857. for (const auto& it : _networks) {
  1858. uint64_t nwid_i = it.first;
  1859. char nwidTmp[64];
  1860. OSUtils::ztsnprintf(nwidTmp, sizeof(nwidTmp), "%.16llx", nwid_i);
  1861. tx.zremrangebyscore("network-nodes-online:{" + controllerId + "}:" + nwidTmp, sw::redis::RightBoundedInterval<double>(expireOld, sw::redis::BoundType::LEFT_OPEN));
  1862. }
  1863. }
  1864. tx.exec();
  1865. fprintf(stderr, "%s: Updated online status of %d members\n", _myAddressStr.c_str(), count);
  1866. return count;
  1867. }
  1868. #endif // ZT_CONTROLLER_USE_LIBPQ