ApsaraMQ for Kafka integrates with HouseKeeing to provide the inspection feature to regularly scan and diagnose clusters and generate alerts for unhealthy instances. This topic describes the inspection feature of ApsaraMQ for Kafka.
Overview
To improve the cluster stability and service reliability of ApsaraMQ for Kafka, the ApsaraMQ for Kafka technical team has optimized the inspection feature.
Inspection on the messaging feature in seconds
You can use the inspection feature to quickly check the messaging feature by using internal topics. If issues, such as messaging failures, increased latency, abnormal throughput, and connection exceptions in dependent components, occur, the system immediately sends alerts to the ApsaraMQ for Kafka technical team. This way, the technical team can quickly respond to and resolve the issues and thus ensure the continuity and stability of the business.
Strict verification on topic offsets and metadata
The inspection feature of ApsaraMQ for Kafka verifies information such as the offsets and metadata of topics. The verification mechanism ensures the correctness and consistency of memory data and prevents potential problems caused by data inconsistency.
Powerful monitoring capabilities for real-time monitoring of process status
If issues, such as I/O hangs, thread deadlocks, JVM crashes, Connection flooding attacks, and memory leaks, occur during system running, the inspection feature can quickly respond to the issues and automatically restart ApsaraMQ for Kafka processes. This ensures the continuity and availability of ApsaraMQ for Kafka and greatly shortens the time window for troubleshooting.
Usage notes
The internal topics used for inspection are
__alikafka_housekeeping_local_topicand__alikafka_housekeeping_cloud_topic.ImportantAfter you deploy an ApsaraMQ for Kafka instance, the system automatically creates the internal topics for inspection.
You can delete internal topics for inspection only after you delete the corresponding instance.
The inspection feature consumes resources. For example, if you use the inspection feature in a cluster with the smallest specification, approximately 100 bits/s of traffic for message production and consumption and 10 MB of storage space are consumed in each topic. The amount of bandwidth consumed positively correlates with the cluster specification.