All Products
Search
Document Center

PolarDB:PolarDB Agent LakeBase monitoring metrics

Last Updated:Jul 13, 2026

PolarDB Agent LakeBase provides 22 monitoring metrics organized into six categories: cluster overview, cache performance, client I/O, client resources, object storage, and metadata engine. These metrics help you understand your file system usage, identify performance bottlenecks, and optimize data access.

View monitoring metrics

  1. Log in to the PolarDB console. In the left-side navigation pane, click Clusters. Select the region where your cluster is located, and then click the target cluster ID to open the cluster details page.

  2. In the left-side navigation pane, choose PolarDB AI > PolarDB AI Overview.

  3. In the AI Cache section, click the ID of the target instance to open the instance details page.

  4. Click the Performance Monitoring tab.

Cluster overview metrics

Cluster overview metrics provide a high-level view of your storage usage and connection status.

Metric

Unit

Description

File and Folder Count

Count

The total number of files and directories stored in the cluster. This value reflects the total number of inodes in the file system, including regular files and directories.

Data Volume

Byte

The total logical storage space occupied by all files in the cluster. This value is the sum of the sizes of all files and does not include file system metadata overhead.

Client Count

Count

The current number of clients (mount points) connected to the cluster.

Cached Data Volume

Byte

The total size of data cached on the cache nodes.

Cache performance metrics

Cache performance metrics reflect the hit efficiency and request activity of the cache layer.

Metric

Unit

Description

Cache Hit Ratio (Bandwidth)

%

The percentage of total read data that is served from the cache nodes.

Cache Hit Ratio (Count)

%

The percentage of total read requests that are served from the cache nodes.

GET Cache Request Count

Count

The total number of Get requests sent to the cache nodes.

PUT Cache Request Count

Count

The total number of Put requests sent to the cache nodes.

Client I/O metrics

Client I/O metrics track client-side read and write activity, including bandwidth and IOPS.

Metric

Unit

Description

Client Write Bandwidth

Byte/s

The throughput of data written by clients.

Client Read Bandwidth

Byte/s

The throughput of data read by clients.

Client Write IOPS

Count/s

The number of write operations initiated by clients per second.

Client Read IOPS

Count/s

The number of read operations initiated by clients per second.

Client GET Bandwidth

Byte/s

The throughput of data retrieved by clients through Get operations.

Client Flush IOPS

Count/s

The number of Flush operations triggered by clients per second.

Client resource metrics

Client resource metrics track the resource consumption of client processes. These metrics help you determine if a client has become a performance bottleneck.

Metric

Unit

Description

Client CPU

%

The client's CPU utilization.

Client Memory

Byte

The amount of memory the client uses.

Object storage metrics

Object storage metrics track data interaction between the AI cache and backend OSS. These metrics reflect the workload generated by cache penetration and data persistence.

Metric

Unit

Description

Cache Node OSS GET Bandwidth

Byte/s

The throughput of data read (Get) from OSS by the cache nodes.

Cache Node OSS GET IOPS

Count/s

The number of read (Get) requests per second sent from the cache nodes to OSS.

OSS PUT Bandwidth

Byte/s

The throughput of data written (Put) to OSS by clients and cache nodes.

OSS PUT IOPS

Count/s

The number of write (Put) requests per second sent to OSS by clients and cache nodes.

OSS Delete IOPS

Count/s

The number of delete (Delete) requests per second sent to OSS by clients and cache nodes.

Metadata engine metrics

Metadata engine metrics track the workload of file system metadata operations.

Metric

Unit

Description

Metadata Engine Operations

Count

The total number of operations initiated by clients to the metadata engine.