librdkafka: 如何设置Kafka消费者订阅消息的起始偏移位置
缺省配置
默认情况下，Kafka消费者从最后一次提交的偏移量位置（offset）开始消费消息，如果Topic+Partition和Group之前没有提交过偏移量，它订阅消息开始位置取决于Topic的配置属性auto.offset.reset的设置。默认为最新(latest)，也就是在分区末尾开始消耗（仅消费新消息）。相关配置可以参考官方文档：https://kafka.apache.org/documentation/#topicconfigs。
方便查阅，截个图：
相关接口信息
librdkafka提供了assign() API，通过设置rd_kafka_topic_partition_t的.offset属性，你可以指定每一个Partition的起始偏移。偏移量可以是一个绝对的偏移(>0)，或逻辑偏移 (BEGINNING, END, STORED, TAIL(…))。
rdkafka.h头文件中定义了Partition的管理结构rd_kafka_topic_partition_t，包含offset信息；同时提供了逻辑偏移的定义RD_KAFKA_OFFSET_XXX。
/**
* @brief Generic place holder for a specific Topic+Partition.
*
* @sa rd_kafka_topic_partition_list_new()
*/
typedef struct rd_kafka_topic_partition_s {
char *topic; /**< Topic name */
int32_t partition; /**< Partition */
int64_t offset; /**< Offset */
void *metadata; /**< Metadata */
size_t metadata_size; /**< Metadata size */
void *opaque; /**< Application opaque */
rd_kafka_resp_err_t err; /**< Error code, depending on use. */
void *_private; /**< INTERNAL USE ONLY,
* INITIALIZE TO ZERO, DO NOT TOUCH */
} rd_kafka_topic_partition_t;
#define RD_KAFKA_OFFSET_BEGINNING -2 /**< Start consuming from beginning of
* kafka partition queue: oldest msg */
#define RD_KAFKA_OFFSET_END -1 /**< Start consuming from end of kafka
* partition queue: next msg */
#define RD_KAFKA_OFFSET_STORED -1000 /**< Start consuming from offset retrieved
* from offset store */
#define RD_KAFKA_OFFSET_INVALID -1001 /**< Invalid offset */
/** @cond NO_DOC */
#define RD_KAFKA_OFFSET_TAIL_BASE -2000 /* internal: do not use */
/** @endcond */
/**
* @brief Start consuming \p CNT messages from topic's current end offset.
*
* That is, if current end offset is 12345 and \p CNT is 200, it will start
* consuming from offset \c 12345-200 = \c 12145. */
#define RD_KAFKA_OFFSET_TAIL(CNT) (RD_KAFKA_OFFSET_TAIL_BASE - (CNT))
通过rd_kafka_assign()函数接口可以配置需要消费的Partition信息。
/**
* @brief Atomic assignment of partitions to consume.
*
* The new \p partitions will replace the existing assignment.
*
* When used from a rebalance callback the application shall pass the
* partition list passed to the callback (or a copy of it) (even if the list
* is empty) rather than NULL to maintain internal join state.
* A zero-length \p partitions will treat the partitions as a valid,
* albeit empty, assignment, and maintain internal state, while a \c NULL
* value for \p partitions will reset and clear the internal state.
*/
RD_EXPORT rd_kafka_resp_err_t
rd_kafka_assign (rd_kafka_t *rk,
const rd_kafka_topic_partition_list_t *partitions);
如何配置offset
对于消费者来说，有两个场景来修改订阅的Parttion offset信息：一是系统初始化时直接指定offset信息，二是消费者群组重平衡(rebalance)的回调函数。接下来分别介绍一下。
系统初始化时，指定offset示例：
rd_kafka_topic_partition_list_t *partitions;
partitions = rd_kafka_topic_partition_list_new(0);
rd_kafka_topic_partition_list_add(partitions, "mytopic", 3)->offset = 1234;
rd_kafka_assign(rk, partitions);
rd_kafka_topic_partition_list_destroy(partitions);
rebalance_cb()回调函数中，指定offset示例。
void my_rebalance_cb (rd_kafka_t *rk, rd_kafka_resp_err_t err,
rd_kafka_topic_partition_list_t *partitions, void *opaque) {
if (err == RD_KAFKA_RESP_ERR__ASSIGN_PARTITIONS) {
rd_kafka_topic_partition_t *part;
if ((part = rd_kafka_topic_partition_list_find(partitions, "mytopic", 3)))
part->offset = 1234;
rd_kafka_assign(rk, partitions);
} else {
rd_kafka_assign(rk, NULL);
}
}
欢迎订阅个人公众号
打个广告，欢迎订阅个人的公众号，文章会在公众号和博客同步发布。
