/* * librdkafka - Apache Kafka C library * * Copyright (c) 2012-2022, 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 "test.h" /* Typical include path would be , but this program * is built from within the librdkafka source tree and thus differs. */ #include "rdkafka.h" /* for Kafka driver */ /** * Consumer: various offset commit constellations, matrix: * enable.auto.commit, enable.auto.offset.store, async */ static char *topic; static const int msgcnt = 100; static const int partition = 0; static uint64_t testid; static int64_t expected_offset = 0; static int64_t committed_offset = -1; static void offset_commit_cb(rd_kafka_t *rk, rd_kafka_resp_err_t err, rd_kafka_topic_partition_list_t *offsets, void *opaque) { rd_kafka_topic_partition_t *rktpar; TEST_SAYL(3, "Offset committed: %s:\n", rd_kafka_err2str(err)); if (err == RD_KAFKA_RESP_ERR__NO_OFFSET) return; test_print_partition_list(offsets); if (err) TEST_FAIL("Offset commit failed: %s", rd_kafka_err2str(err)); if (offsets->cnt == 0) TEST_FAIL( "Expected at least one partition in offset_commit_cb"); /* Find correct partition */ if (!(rktpar = rd_kafka_topic_partition_list_find(offsets, topic, partition))) return; if (rktpar->err) TEST_FAIL("Offset commit failed for partitioń : %s", rd_kafka_err2str(rktpar->err)); if (rktpar->offset > expected_offset) TEST_FAIL("Offset committed %" PRId64 " > expected offset %" PRId64, rktpar->offset, expected_offset); if (rktpar->offset < committed_offset) TEST_FAIL("Old offset %" PRId64 " (re)committed: " "should be above committed_offset %" PRId64, rktpar->offset, committed_offset); else if (rktpar->offset == committed_offset) TEST_SAYL(1, "Current offset re-committed: %" PRId64 "\n", rktpar->offset); else committed_offset = rktpar->offset; if (rktpar->offset < expected_offset) { TEST_SAYL(3, "Offset committed %" PRId64 " < expected offset %" PRId64 "\n", rktpar->offset, expected_offset); return; } TEST_SAYL(3, "Expected offset committed: %" PRId64 "\n", rktpar->offset); } static void do_offset_test(const char *what, int auto_commit, int auto_store, int async, int subscribe) { test_timing_t t_all; char groupid[64]; rd_kafka_t *rk; rd_kafka_conf_t *conf; rd_kafka_topic_conf_t *tconf; int cnt = 0; const int extra_cnt = 5; rd_kafka_resp_err_t err; rd_kafka_topic_partition_list_t *parts; rd_kafka_topic_partition_t *rktpar; int64_t next_offset = -1; SUB_TEST_QUICK("%s", what); test_conf_init(&conf, &tconf, subscribe ? 30 : 10); test_conf_set(conf, "session.timeout.ms", "6000"); test_conf_set(conf, "enable.auto.commit", auto_commit ? "true" : "false"); test_conf_set(conf, "enable.auto.offset.store", auto_store ? "true" : "false"); test_conf_set(conf, "auto.commit.interval.ms", "500"); rd_kafka_conf_set_offset_commit_cb(conf, offset_commit_cb); test_topic_conf_set(tconf, "auto.offset.reset", "smallest"); test_str_id_generate(groupid, sizeof(groupid)); test_conf_set(conf, "group.id", groupid); rd_kafka_conf_set_default_topic_conf(conf, tconf); TIMING_START(&t_all, "%s", what); expected_offset = 0; committed_offset = -1; /* MO: * - Create consumer. * - Start consuming from beginning * - Perform store & commits according to settings * - Stop storing&committing when half of the messages are consumed, * - but consume 5 more to check against. * - Query position. * - Destroy consumer. * - Create new consumer with same group.id using stored offsets * - Should consume the expected message. */ /* Create kafka instance */ rk = test_create_handle(RD_KAFKA_CONSUMER, rd_kafka_conf_dup(conf)); rd_kafka_poll_set_consumer(rk); if (subscribe) { test_consumer_subscribe(rk, topic); } else { parts = rd_kafka_topic_partition_list_new(1); rd_kafka_topic_partition_list_add(parts, topic, partition); test_consumer_assign("ASSIGN", rk, parts); rd_kafka_topic_partition_list_destroy(parts); } while (cnt - extra_cnt < msgcnt / 2) { rd_kafka_message_t *rkm; rkm = rd_kafka_consumer_poll(rk, 10 * 1000); if (!rkm) continue; if (rkm->err == RD_KAFKA_RESP_ERR__TIMED_OUT) TEST_FAIL("%s: Timed out waiting for message %d", what, cnt); else if (rkm->err) TEST_FAIL("%s: Consumer error: %s", what, rd_kafka_message_errstr(rkm)); /* Offset of next message. */ next_offset = rkm->offset + 1; if (cnt < msgcnt / 2) { if (!auto_store) { err = rd_kafka_offset_store( rkm->rkt, rkm->partition, rkm->offset); if (err) TEST_FAIL( "%s: offset_store failed: %s\n", what, rd_kafka_err2str(err)); } expected_offset = rkm->offset + 1; if (!auto_commit) { test_timing_t t_commit; TIMING_START(&t_commit, "%s @ %" PRId64, async ? "commit.async" : "commit.sync", rkm->offset + 1); err = rd_kafka_commit_message(rk, rkm, async); TIMING_STOP(&t_commit); if (err) TEST_FAIL("%s: commit failed: %s\n", what, rd_kafka_err2str(err)); } } else if (auto_store && auto_commit) expected_offset = rkm->offset + 1; rd_kafka_message_destroy(rkm); cnt++; } TEST_SAY("%s: done consuming after %d messages, at offset %" PRId64 ", next_offset %" PRId64 "\n", what, cnt, expected_offset, next_offset); if ((err = rd_kafka_assignment(rk, &parts))) TEST_FAIL("%s: failed to get assignment(): %s\n", what, rd_kafka_err2str(err)); /* Verify position */ if ((err = rd_kafka_position(rk, parts))) TEST_FAIL("%s: failed to get position(): %s\n", what, rd_kafka_err2str(err)); if (!(rktpar = rd_kafka_topic_partition_list_find(parts, topic, partition))) TEST_FAIL("%s: position(): topic lost\n", what); if (rktpar->offset != next_offset) TEST_FAIL("%s: Expected position() offset %" PRId64 ", got %" PRId64, what, next_offset, rktpar->offset); TEST_SAY("%s: Position is at %" PRId64 ", good!\n", what, rktpar->offset); /* Pause messages while waiting so we can serve callbacks * without having more messages received. */ if ((err = rd_kafka_pause_partitions(rk, parts))) TEST_FAIL("%s: failed to pause partitions: %s\n", what, rd_kafka_err2str(err)); rd_kafka_topic_partition_list_destroy(parts); /* Fire off any enqueued offset_commit_cb */ test_consumer_poll_no_msgs(what, rk, testid, 0); TEST_SAY("%s: committed_offset %" PRId64 ", expected_offset %" PRId64 "\n", what, committed_offset, expected_offset); if (!auto_commit && !async) { /* Sync commits should be up to date at this point. */ if (committed_offset != expected_offset) TEST_FAIL("%s: Sync commit: committed offset %" PRId64 " should be same as expected offset " "%" PRId64, what, committed_offset, expected_offset); } else { /* Wait for offset commits to catch up */ while (committed_offset < expected_offset) { TEST_SAYL(2, "%s: Wait for committed offset %" PRId64 " to reach expected offset %" PRId64 "\n", what, committed_offset, expected_offset); test_consumer_poll_no_msgs(what, rk, testid, 1000); } } TEST_SAY( "%s: phase 1 complete, %d messages consumed, " "next expected offset is %" PRId64 "\n", what, cnt, expected_offset); /* Issue #827: cause committed() to return prematurely by specifying * low timeout. The bug (use after free) will only * be catched by valgrind. * * rusage: this triggers a bunch of protocol requests which * increase .ucpu, .scpu, .ctxsw. */ do { parts = rd_kafka_topic_partition_list_new(1); rd_kafka_topic_partition_list_add(parts, topic, partition); err = rd_kafka_committed(rk, parts, 1); rd_kafka_topic_partition_list_destroy(parts); if (err) TEST_SAY("Issue #827: committed() returned %s\n", rd_kafka_err2str(err)); } while (err != RD_KAFKA_RESP_ERR__TIMED_OUT); /* Query position */ parts = rd_kafka_topic_partition_list_new(1); rd_kafka_topic_partition_list_add(parts, topic, partition); err = rd_kafka_committed(rk, parts, tmout_multip(5 * 1000)); if (err) TEST_FAIL("%s: committed() failed: %s", what, rd_kafka_err2str(err)); if (!(rktpar = rd_kafka_topic_partition_list_find(parts, topic, partition))) TEST_FAIL("%s: committed(): topic lost\n", what); if (rktpar->offset != expected_offset) TEST_FAIL("%s: Expected committed() offset %" PRId64 ", got %" PRId64, what, expected_offset, rktpar->offset); TEST_SAY("%s: Committed offset is at %" PRId64 ", good!\n", what, rktpar->offset); rd_kafka_topic_partition_list_destroy(parts); test_consumer_close(rk); rd_kafka_destroy(rk); /* Fire up a new consumer and continue from where we left off. */ TEST_SAY("%s: phase 2: starting new consumer to resume consumption\n", what); rk = test_create_handle(RD_KAFKA_CONSUMER, conf); rd_kafka_poll_set_consumer(rk); if (subscribe) { test_consumer_subscribe(rk, topic); } else { parts = rd_kafka_topic_partition_list_new(1); rd_kafka_topic_partition_list_add(parts, topic, partition); test_consumer_assign("ASSIGN", rk, parts); rd_kafka_topic_partition_list_destroy(parts); } while (cnt < msgcnt) { rd_kafka_message_t *rkm; rkm = rd_kafka_consumer_poll(rk, 10 * 1000); if (!rkm) continue; if (rkm->err == RD_KAFKA_RESP_ERR__TIMED_OUT) TEST_FAIL("%s: Timed out waiting for message %d", what, cnt); else if (rkm->err) TEST_FAIL("%s: Consumer error: %s", what, rd_kafka_message_errstr(rkm)); if (rkm->offset != expected_offset) TEST_FAIL("%s: Received message offset %" PRId64 ", expected %" PRId64 " at msgcnt %d/%d\n", what, rkm->offset, expected_offset, cnt, msgcnt); rd_kafka_message_destroy(rkm); expected_offset++; cnt++; } TEST_SAY("%s: phase 2: complete\n", what); test_consumer_close(rk); rd_kafka_destroy(rk); TIMING_STOP(&t_all); SUB_TEST_PASS(); } static void empty_offset_commit_cb(rd_kafka_t *rk, rd_kafka_resp_err_t err, rd_kafka_topic_partition_list_t *offsets, void *opaque) { rd_kafka_resp_err_t expected = *(rd_kafka_resp_err_t *)opaque; int valid_offsets = 0; int i; TEST_SAY( "Offset commit callback for %d partitions: %s (expecting %s)\n", offsets ? offsets->cnt : 0, rd_kafka_err2str(err), rd_kafka_err2str(expected)); if (expected != err) TEST_FAIL("Offset commit cb: expected %s, got %s", rd_kafka_err2str(expected), rd_kafka_err2str(err)); for (i = 0; i < offsets->cnt; i++) { TEST_SAY("committed: %s [%" PRId32 "] offset %" PRId64 ": %s\n", offsets->elems[i].topic, offsets->elems[i].partition, offsets->elems[i].offset, rd_kafka_err2str(offsets->elems[i].err)); if (expected == RD_KAFKA_RESP_ERR_NO_ERROR) TEST_ASSERT(offsets->elems[i].err == expected); if (offsets->elems[i].offset > 0) valid_offsets++; } if (expected == RD_KAFKA_RESP_ERR_NO_ERROR) { /* If no error is expected we instead expect one proper offset * to have been committed. */ TEST_ASSERT(valid_offsets > 0); } } /** * Trigger an empty cgrp commit (issue #803) */ static void do_empty_commit(void) { rd_kafka_t *rk; char group_id[64]; rd_kafka_conf_t *conf; rd_kafka_topic_conf_t *tconf; rd_kafka_resp_err_t err, expect; SUB_TEST_QUICK(); test_conf_init(&conf, &tconf, 20); test_conf_set(conf, "enable.auto.commit", "false"); test_topic_conf_set(tconf, "auto.offset.reset", "earliest"); test_str_id_generate(group_id, sizeof(group_id)); TEST_SAY(_C_MAG "[ do_empty_commit group.id %s ]\n", group_id); rk = test_create_consumer(group_id, NULL, conf, tconf); test_consumer_subscribe(rk, topic); test_consumer_poll("consume", rk, testid, -1, -1, 100, NULL); TEST_SAY("First commit\n"); expect = RD_KAFKA_RESP_ERR_NO_ERROR; err = rd_kafka_commit_queue(rk, NULL, NULL, empty_offset_commit_cb, &expect); if (err != expect) TEST_FAIL("commit failed: %s", rd_kafka_err2str(err)); else TEST_SAY("First commit returned %s\n", rd_kafka_err2str(err)); TEST_SAY("Second commit, should be empty\n"); expect = RD_KAFKA_RESP_ERR__NO_OFFSET; err = rd_kafka_commit_queue(rk, NULL, NULL, empty_offset_commit_cb, &expect); if (err != RD_KAFKA_RESP_ERR__NO_OFFSET) TEST_FAIL("unexpected commit result, wanted NO_OFFSET, got: %s", rd_kafka_err2str(err)); else TEST_SAY("Second commit returned %s\n", rd_kafka_err2str(err)); test_consumer_close(rk); rd_kafka_destroy(rk); SUB_TEST_PASS(); } /** * Commit non-existent topic (issue #704) */ static void nonexist_offset_commit_cb(rd_kafka_t *rk, rd_kafka_resp_err_t err, rd_kafka_topic_partition_list_t *offsets, void *opaque) { int i; int failed_offsets = 0; TEST_SAY("Offset commit callback for %d partitions: %s\n", offsets ? offsets->cnt : 0, rd_kafka_err2str(err)); TEST_ASSERT(offsets != NULL); for (i = 0; i < offsets->cnt; i++) { TEST_SAY("committed: %s [%" PRId32 "] offset %" PRId64 ": %s\n", offsets->elems[i].topic, offsets->elems[i].partition, offsets->elems[i].offset, rd_kafka_err2str(offsets->elems[i].err)); failed_offsets += offsets->elems[i].err ? 1 : 0; } TEST_ASSERT(err == RD_KAFKA_RESP_ERR_UNKNOWN_TOPIC_OR_PART, "expected unknown Topic or partition, not %s", rd_kafka_err2str(err)); TEST_ASSERT(offsets->cnt == 2, "expected %d offsets", offsets->cnt); TEST_ASSERT(failed_offsets == offsets->cnt, "expected %d offsets to have failed, got %d", offsets->cnt, failed_offsets); } static void do_nonexist_commit(void) { rd_kafka_t *rk; char group_id[64]; rd_kafka_conf_t *conf; rd_kafka_topic_conf_t *tconf; rd_kafka_topic_partition_list_t *offsets; const char *unk_topic = test_mk_topic_name(__FUNCTION__, 1); rd_kafka_resp_err_t err; SUB_TEST_QUICK(); test_conf_init(&conf, &tconf, 20); /* Offset commit deferrals when the broker is down is limited to * session.timeout.ms. With 0.9 brokers and api.version.request=true * the initial connect to all brokers will take 10*2 seconds * and the commit_queue() below will time out too quickly. * Set the session timeout high here to avoid it. */ test_conf_set(conf, "session.timeout.ms", "60000"); test_str_id_generate(group_id, sizeof(group_id)); test_conf_set(conf, "group.id", group_id); rd_kafka_conf_set_default_topic_conf(conf, tconf); TEST_SAY(_C_MAG "[ do_nonexist_commit group.id %s ]\n", group_id); rk = test_create_handle(RD_KAFKA_CONSUMER, conf); rd_kafka_poll_set_consumer(rk); TEST_SAY("Try nonexist commit\n"); offsets = rd_kafka_topic_partition_list_new(2); rd_kafka_topic_partition_list_add(offsets, unk_topic, 0)->offset = 123; rd_kafka_topic_partition_list_add(offsets, unk_topic, 1)->offset = 456; err = rd_kafka_commit_queue(rk, offsets, NULL, nonexist_offset_commit_cb, NULL); TEST_SAY("nonexist commit returned %s\n", rd_kafka_err2str(err)); if (err != RD_KAFKA_RESP_ERR_UNKNOWN_TOPIC_OR_PART) TEST_FAIL("commit() should give UnknownTopicOrPart, not: %s", rd_kafka_err2str(err)); rd_kafka_topic_partition_list_destroy(offsets); test_consumer_close(rk); rd_kafka_destroy(rk); SUB_TEST_PASS(); } int main_0030_offset_commit(int argc, char **argv) { topic = rd_strdup(test_mk_topic_name(__FUNCTION__, 1)); testid = test_produce_msgs_easy(topic, 0, partition, msgcnt); do_empty_commit(); do_nonexist_commit(); do_offset_test("AUTO.COMMIT & AUTO.STORE", 1 /* enable.auto.commit */, 1 /* enable.auto.offset.store */, 0 /* not used. */, 1 /* use subscribe */); do_offset_test("MANUAL.COMMIT.ASYNC & AUTO.STORE", 0 /* enable.auto.commit */, 1 /* enable.auto.offset.store */, 1 /* async */, 1 /* use subscribe */); do_offset_test("AUTO.COMMIT.ASYNC & AUTO.STORE & ASSIGN", 1 /* enable.auto.commit */, 1 /* enable.auto.offset.store */, 0 /* not used. */, 0 /* use assign */); if (test_quick) { rd_free(topic); return 0; } do_offset_test("AUTO.COMMIT & MANUAL.STORE", 1 /* enable.auto.commit */, 0 /* enable.auto.offset.store */, 0 /* not used */, 1 /* use subscribe */); do_offset_test("MANUAL.COMMIT.SYNC & AUTO.STORE", 0 /* enable.auto.commit */, 1 /* enable.auto.offset.store */, 0 /* async */, 1 /* use subscribe */); do_offset_test("MANUAL.COMMIT.ASYNC & MANUAL.STORE", 0 /* enable.auto.commit */, 0 /* enable.auto.offset.store */, 1 /* sync */, 1 /* use subscribe */); do_offset_test("MANUAL.COMMIT.SYNC & MANUAL.STORE", 0 /* enable.auto.commit */, 0 /* enable.auto.offset.store */, 0 /* sync */, 1 /* use subscribe */); rd_free(topic); return 0; }