/* * librdkafka - Apache Kafka C library * * Copyright (c) 2019, Magnus Edenhill * All rights reserved. * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that the following conditions are met: * * 1. Redistributions of source code must retain the above copyright notice, * this list of conditions and the following disclaimer. * 2. Redistributions in binary form must reproduce the above copyright notice, * this list of conditions and the following disclaimer in the documentation * and/or other materials provided with the distribution. * * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE * ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE * POSSIBILITY OF SUCH DAMAGE. */ #include "testcpp.h" #if WITH_RAPIDJSON #include #include #include #include #include #include #include #include #include "rdkafka.h" #include #include #include #include #include #include /** * @brief A basic test of fetch from follower funtionality * - produces a bunch of messages to a replicated topic. * - configure the consumer such that `client.rack` is different from the * broker's `broker.rack` (and use * org.apache.kafka.common.replica.RackAwareReplicaSelector). * - consume the messages, and check they are as expected. * - use rxbytes from the statistics event to confirm that * the messages were retrieved from the replica broker (not the * leader). */ #define test_assert(cond, msg) \ do { \ if (!(cond)) \ Test::Say(msg); \ } while (0) class TestEvent2Cb : public RdKafka::EventCb { public: static bool should_capture_stats; static bool has_captured_stats; static std::map rxbytes; void event_cb(RdKafka::Event &event) { switch (event.type()) { case RdKafka::Event::EVENT_LOG: Test::Say(event.str() + "\n"); break; case RdKafka::Event::EVENT_STATS: if (should_capture_stats) { rapidjson::Document d; if (d.Parse(event.str().c_str()).HasParseError()) Test::Fail(tostr() << "Failed to parse stats JSON: " << rapidjson::GetParseError_En(d.GetParseError()) << " at " << d.GetErrorOffset()); /* iterate over brokers. */ rapidjson::Pointer jpath((const char *)"/brokers"); rapidjson::Value *pp = rapidjson::GetValueByPointer(d, jpath); if (pp == NULL) return; for (rapidjson::Value::ConstMemberIterator itr = pp->MemberBegin(); itr != pp->MemberEnd(); ++itr) { std::string broker_name = itr->name.GetString(); size_t broker_id_idx = broker_name.rfind('/'); if (broker_id_idx == (size_t)-1) continue; std::string broker_id = broker_name.substr( broker_id_idx + 1, broker_name.size() - broker_id_idx - 1); int64_t broker_rxbytes = itr->value.FindMember("rxbytes")->value.GetInt64(); rxbytes[atoi(broker_id.c_str())] = broker_rxbytes; } has_captured_stats = true; break; } default: break; } } }; bool TestEvent2Cb::should_capture_stats; bool TestEvent2Cb::has_captured_stats; std::map TestEvent2Cb::rxbytes; static TestEvent2Cb ex_event_cb; static void get_brokers_info(std::string &topic_str, int32_t *leader, std::vector &brokers) { std::string errstr; RdKafka::ErrorCode err; class RdKafka::Metadata *metadata; /* Determine the ids of the brokers that the partition has replicas * on and which one of those is the leader. */ RdKafka::Conf *pConf; Test::conf_init(&pConf, NULL, 10); RdKafka::Producer *p = RdKafka::Producer::create(pConf, errstr); delete pConf; test_assert(p, tostr() << "Failed to create producer: " << errstr); RdKafka::Topic *topic = RdKafka::Topic::create(p, topic_str, NULL, errstr); test_assert(topic, tostr() << "Failed to create topic: " << errstr); err = p->metadata(0, topic, &metadata, tmout_multip(5000)); test_assert( err == RdKafka::ERR_NO_ERROR, tostr() << "%% Failed to acquire metadata: " << RdKafka::err2str(err)); test_assert(metadata->topics()->size() == 1, tostr() << "expecting metadata for exactly one topic. " << "have metadata for " << metadata->topics()->size() << "topics"); RdKafka::Metadata::TopicMetadataIterator topicMetadata = metadata->topics()->begin(); RdKafka::TopicMetadata::PartitionMetadataIterator partitionMetadata = (*topicMetadata)->partitions()->begin(); *leader = (*partitionMetadata)->leader(); size_t idx = 0; RdKafka::PartitionMetadata::ReplicasIterator replicasIterator; for (replicasIterator = (*partitionMetadata)->replicas()->begin(); replicasIterator != (*partitionMetadata)->replicas()->end(); ++replicasIterator) { brokers.push_back(*replicasIterator); idx++; } delete metadata; delete topic; delete p; } /** * @brief Wait for up to \p tmout for any type of admin result. * @returns the event */ rd_kafka_event_t *test_wait_admin_result(rd_kafka_queue_t *q, rd_kafka_event_type_t evtype, int tmout) { rd_kafka_event_t *rkev; while (1) { rkev = rd_kafka_queue_poll(q, tmout); if (!rkev) Test::Fail(tostr() << "Timed out waiting for admin result (" << evtype << ")\n"); if (rd_kafka_event_type(rkev) == evtype) return rkev; if (rd_kafka_event_type(rkev) == RD_KAFKA_EVENT_ERROR) { Test::Say(tostr() << "Received error event while waiting for " << evtype << ": " << rd_kafka_event_error_string(rkev) << ": ignoring"); continue; } test_assert(rd_kafka_event_type(rkev) == evtype, tostr() << "Expected event type " << evtype << ", got " << rd_kafka_event_type(rkev) << " (" << rd_kafka_event_name(rkev) << ")"); } return NULL; } /** * @returns the number of broker.rack values configured across all brokers. */ static int get_broker_rack_count(std::vector &replica_ids) { std::string errstr; RdKafka::Conf *pConf; Test::conf_init(&pConf, NULL, 10); RdKafka::Producer *p = RdKafka::Producer::create(pConf, errstr); delete pConf; rd_kafka_queue_t *mainq = rd_kafka_queue_get_main(p->c_ptr()); std::set racks; for (size_t i = 0; i < replica_ids.size(); ++i) { std::string name = tostr() << replica_ids[i]; rd_kafka_ConfigResource_t *config = rd_kafka_ConfigResource_new(RD_KAFKA_RESOURCE_BROKER, &name[0]); rd_kafka_AdminOptions_t *options; char cerrstr[128]; options = rd_kafka_AdminOptions_new(p->c_ptr(), RD_KAFKA_ADMIN_OP_ANY); rd_kafka_resp_err_t err = rd_kafka_AdminOptions_set_request_timeout( options, 10000, cerrstr, sizeof(cerrstr)); test_assert(!err, cerrstr); rd_kafka_DescribeConfigs(p->c_ptr(), &config, 1, options, mainq); rd_kafka_ConfigResource_destroy(config); rd_kafka_AdminOptions_destroy(options); rd_kafka_event_t *rkev = test_wait_admin_result( mainq, RD_KAFKA_EVENT_DESCRIBECONFIGS_RESULT, 5000); const rd_kafka_DescribeConfigs_result_t *res = rd_kafka_event_DescribeConfigs_result(rkev); test_assert(res, "expecting describe config results to be not NULL"); err = rd_kafka_event_error(rkev); const char *errstr2 = rd_kafka_event_error_string(rkev); test_assert(!err, tostr() << "Expected success, not " << rd_kafka_err2name(err) << ": " << errstr2); size_t rconfig_cnt; const rd_kafka_ConfigResource_t **rconfigs = rd_kafka_DescribeConfigs_result_resources(res, &rconfig_cnt); test_assert(rconfig_cnt == 1, tostr() << "Expecting 1 resource, got " << rconfig_cnt); err = rd_kafka_ConfigResource_error(rconfigs[0]); errstr2 = rd_kafka_ConfigResource_error_string(rconfigs[0]); size_t entry_cnt; const rd_kafka_ConfigEntry_t **entries = rd_kafka_ConfigResource_configs(rconfigs[0], &entry_cnt); for (size_t j = 0; j < entry_cnt; ++j) { const rd_kafka_ConfigEntry_t *e = entries[j]; const char *cname = rd_kafka_ConfigEntry_name(e); if (!strcmp(cname, "broker.rack")) { const char *val = rd_kafka_ConfigEntry_value(e) ? rd_kafka_ConfigEntry_value(e) : "(NULL)"; racks.insert(std::string(val)); } } rd_kafka_event_destroy(rkev); } rd_kafka_queue_destroy(mainq); delete p; return (int)racks.size(); } static void do_fff_test(void) { /* Produce some messages to a single partition topic * with 3 replicas. */ int msgcnt = 1000; const int msgsize = 100; std::string topic_str = Test::mk_topic_name("0101-fetch-from-follower", 1); test_create_topic(NULL, topic_str.c_str(), 1, 3); test_produce_msgs_easy_size(topic_str.c_str(), 0, 0, msgcnt, msgsize); int leader_id; std::vector replica_ids; get_brokers_info(topic_str, &leader_id, replica_ids); test_assert(replica_ids.size() == 3, tostr() << "expecting three replicas, but " << replica_ids.size() << " were reported."); Test::Say(tostr() << topic_str << " leader id: " << leader_id << ", all replica ids: [" << replica_ids[0] << ", " << replica_ids[1] << ", " << replica_ids[2] << "]\n"); if (get_broker_rack_count(replica_ids) != 3) { Test::Skip("unexpected broker.rack configuration: skipping test.\n"); return; } /* arrange for the consumer's client.rack to align with a broker that is not * the leader. */ int client_rack_id = -1; size_t i; for (i = 0; i < replica_ids.size(); ++i) { if (replica_ids[i] != leader_id) { client_rack_id = replica_ids[i]; break; } } std::string client_rack = tostr() << "RACK" << client_rack_id; Test::Say("client.rack: " + client_rack + "\n"); std::string errstr; RdKafka::Conf *conf; Test::conf_init(&conf, NULL, 10); Test::conf_set(conf, "group.id", topic_str); Test::conf_set(conf, "auto.offset.reset", "earliest"); Test::conf_set(conf, "enable.auto.commit", "false"); Test::conf_set(conf, "statistics.interval.ms", "1000"); conf->set("event_cb", &ex_event_cb, errstr); Test::conf_set(conf, "client.rack", client_rack); RdKafka::KafkaConsumer *c = RdKafka::KafkaConsumer::create(conf, errstr); test_assert(c, "Failed to create KafkaConsumer: " + errstr); delete conf; /* Subscribe */ std::vector topics; topics.push_back(topic_str); RdKafka::ErrorCode err; if ((err = c->subscribe(topics))) Test::Fail("subscribe failed: " + RdKafka::err2str(err)); /* Start consuming */ Test::Say("Consuming topic " + topic_str + "\n"); int cnt = 0; while (cnt < msgcnt) { RdKafka::Message *msg = c->consume(tmout_multip(1000)); switch (msg->err()) { case RdKafka::ERR__TIMED_OUT: break; case RdKafka::ERR_NO_ERROR: { test_assert(msg->len() == 100, "expecting message value size to be 100"); char *cnt_str_start_ptr = strstr((char *)msg->payload(), "msg=") + 4; test_assert(cnt_str_start_ptr, "expecting 'msg=' in message payload"); char *cnt_str_end_ptr = strstr(cnt_str_start_ptr, "\n"); test_assert(cnt_str_start_ptr, "expecting '\n' following 'msg=' in message payload"); *cnt_str_end_ptr = '\0'; int msg_cnt = atoi(cnt_str_start_ptr); test_assert(msg_cnt == cnt, "message consumed out of order"); cnt++; } break; default: Test::Fail("Consume error: " + msg->errstr()); break; } delete msg; } /* rely on the test timeout to prevent an infinite loop in * the (unlikely) event that the statistics callback isn't * called. */ Test::Say("Capturing rxbytes statistics\n"); TestEvent2Cb::should_capture_stats = true; while (!TestEvent2Cb::has_captured_stats) { RdKafka::Message *msg = c->consume(tmout_multip(500)); delete msg; } for (i = 0; i < replica_ids.size(); ++i) Test::Say( tostr() << _C_YEL << "rxbytes for replica on broker " << replica_ids[i] << ": " << TestEvent2Cb::rxbytes[replica_ids[i]] << (replica_ids[i] == leader_id ? " (leader)" : "") << (replica_ids[i] == client_rack_id ? " (preferred replica)" : "") << "\n"); for (i = 0; i < replica_ids.size(); ++i) if (replica_ids[i] != client_rack_id) test_assert( TestEvent2Cb::rxbytes[replica_ids[i]] < TestEvent2Cb::rxbytes[client_rack_id], "rxbytes was not highest on broker corresponding to client.rack."); test_assert( TestEvent2Cb::rxbytes[client_rack_id] > msgcnt * msgsize, tostr() << "expecting rxbytes of client.rack broker to be at least " << msgcnt * msgsize << " but it was " << TestEvent2Cb::rxbytes[client_rack_id]); Test::Say("Done\n"); // Manual test 1: // - change the lease period from 5 minutes to 5 seconds (modify // rdkafka_partition.c) // - change the max lease grant period from 1 minute to 10 seconds (modify // rdkafka_broker.c) // - add infinite consume loop to the end of this test. // - observe: // - the partition gets delegated to the preferred replica. // - the messages get consumed. // - the lease expires. // - the partition is reverted to the leader. // - the toppar is backed off, and debug message noting the faster than // expected delegation to a replica. // Manual test 2: // - same modifications as above. // - add Test::conf_set(conf, "topic.metadata.refresh.interval.ms", "3000"); // - observe: // - that metadata being periodically received and not interfering with // anything. c->close(); delete c; } #endif extern "C" { int main_0101_fetch_from_follower(int argc, char **argv) { #if WITH_RAPIDJSON do_fff_test(); #else Test::Skip("RapidJSON >=1.1.0 not available\n"); #endif return 0; } }