View monitoring information

Prerequisites

You have created the ApsaraDB for SelectDB service-linked role for Alibaba Cloud SelectDB. By default, this role includes permissions to access Application Real-Time Monitoring Service (ARMS). For more information, see Service-linked roles.
To report monitoring data from your Alibaba Cloud SelectDB instance to your own ARMS, you must first activate ARMS. For more information, see Activate ARMS.

Procedure

Log on to the Alibaba Cloud SelectDB console.
In the upper-left corner of the page, select the region where the instance is located.
On the Instances page, click the target Instance ID to open the Instance Details page.
In the left-side navigation pane, click Monitoring and Alerts.
On the Monitoring and Alerts page, click the Monitor tab.
On the Monitor tab, view the monitoring information for the instance. A non-Serverless version supports multiple clusters. You can click Cluster in the upper-left corner of the tab to switch between clusters and view their respective monitoring information.
Note
- Monitoring metrics are displayed on a Grafana dashboard using preset PromQL queries. On the edit page, you can modify PromQL queries to view different monitoring metrics, but your changes cannot be saved.
- By default, the monitoring dashboard displays cluster monitoring information from the last 15 minutes. You can adjust the time range in the upper-right corner of the dashboard.
- By default, the monitoring dashboard does not automatically refresh. You can click the refresh icon in the upper-right corner to manually refresh the data or set an automatic refresh frequency.

Metrics

ApsaraDB for SelectDB provides metrics at three levels: basic node metrics, cluster request metrics, and instance status metrics.

Basic node metrics

Basic metrics for all nodes in an ApsaraDB for SelectDB cluster.

Metric	Description	Unit
CPU usage	The CPU usage of all nodes in the cluster. If the CPU usage remains high, scale up the cluster's compute resources. For more information, see cluster scaling.	%
memory usage	The memory usage of all nodes in the cluster. If the memory usage remains high, scale up the cluster's compute resources. For more information, see cluster scaling.	%
PodNetworkFlowRx	The average rate at which data is written to all nodes in the cluster over the network.	Bytes/s
PodNetworkFlowTx	The average rate at which data is read from all nodes in the cluster over the network. If the query speed for network reads is slow, we recommend that you modify the cache settings to minimize network reads during queries. For more information, see Data caching.	Bytes/s
disk IOPS	The number of disk `I/O` operations per second (IOPS) for all nodes in the cluster. This metric includes two curves: reads and writes. Every 100 GB of disk cache space provides approximately 5,000 IOPS. If you require higher disk IOPS, we recommend that you expand the cluster's cache space. For more information, see cluster scaling.	ops/s
disk bandwidth	The disk bandwidth usage of all nodes in the cluster. This metric includes two curves: read bandwidth and write bandwidth.	Bytes/s

Cluster request metrics

Request metrics at the ApsaraDB for SelectDB cluster level.

Metric	Description	Unit
queries per second (QPS)	The total number of query requests connected to the cluster per second.	Count
ClusterQuerySuccessRate	Formula: ClusterQuerySuccessRate = `Successful queries per minute in the cluster / Total queries per minute in the cluster`. An abnormal drop in the query success rate may indicate a cluster or node failure.	%
Query rate	The average time consumed by all queries in the cluster per minute.	ms
99th query time	The response time at the 99th percentile of requests per minute in the cluster. This metric helps identify slow queries.	ms
ClusterDataLoadRate	The number of data rows imported into the cluster per second.	Rows/s
import data volume	The amount of data imported into the cluster per second.	Bytes/s
cache hit ratio	Formula: Cache hit ratio = `Number of read/write operations that hit the cache in the cluster / Total read/write operations in the cluster`. If the cache hit ratio is too low, consider changing the caching policy or increasing the cache disk size to improve the cache hit ratio. For more information about data caching, see Data caching.	%
cache usage	Formula: Cache usage = `Total cached data in the cluster / Total cache size of the cluster`. Important This metric is supported starting from V4.0.4.	%
remote storage read/write QPS	The QPS for cluster read and write operations on object storage. This metric includes two curves: GET QPS and PUT QPS.	ops/s
remote storage read/write bandwidth	The bandwidth for cluster read and write operations on object storage. This metric includes read bandwidth and write bandwidth.	Bytes/s
compaction score	The compaction pressure of data files in the cluster. This metric chart includes two curves: `Cumulative Score` and `Base Score`. A higher score indicates greater compaction pressure. A score greater than 2000 triggers write backpressure, which can cause data writes to fail for tables with high compaction pressure.	None
ClusterFailNodeCount	The number of failed nodes in the cluster.	Count

Instance status metrics

Status metrics at the ApsaraDB for SelectDB instance level.

Metric	Description	Unit
FE CPU usage	The CPU usage of all FE nodes in the instance. This metric includes two curves: average (avg) and maximum (max).	%
FE JVM memory usage	The JVM memory usage of all FE nodes in the instance. This metric includes two curves: average (avg) and maximum (max).	%
number of Stream Load import jobs	The number of Stream Load import jobs in the instance. For more information about Stream Load, see Stream Load.	Count
average Stream Load job duration	The average duration of Stream Load import jobs in the instance.	ms
Job Import Frequency in Insert Load Mode	The number of Insert Load import jobs that the instance can process per second. For more information about Insert Load, see Insert Into.	jobs/s
Number of Import Jobs in Broker Load Mode	The number of Broker Load import jobs in each status in the instance. For more information about Broker Load, see Broker Load.	Count
Number of Import Jobs in Routine Load Mode	The number of Routine Load import jobs in each status in the instance. For more information about Routine Load, see Routine Load.	Count
Routine Load import data speed	The number of data rows imported by using Routine Load in the instance per second. For more information about Routine Load, see Routine Load.	Rows/s
Routine Load row error rate	The percentage of errored rows in the total number of rows imported by Routine Load jobs in the instance. For more information about Routine Load, see Routine Load.	%
number of import transactions	The number of import transactions in the instance.	Count
total data size	The total amount of data in the instance.	Byte
User Connection Count	The total number of user connections to the instance, aggregated from all clusters.	Count