CMS and OB Cloud monitoring metric compatibility|OceanBase Cloud |docs|OceanBase Documentations

This topic describes metric mappings and differences in definitions and units to consider when moving from CMS metrics for OceanBase Cloud to OB Cloud monitoring.

Scope

This topic applies to users who already monitor OceanBase Cloud with CMS and need to query metrics, configure charts, or set up alerts in OB Cloud.

CMS and OB Cloud differ in their monitoring data sources, the resources they measure, aggregation methods, and units. Matching metric names do not mean that you can reuse existing queries, chart configurations, or alert thresholds without adjustment. This topic highlights differences to check; it is not a complete metric catalog.

Considerations

Before migrating queries, charts, and alerts, check which resources each metric measures, how it is aggregated, and its unit. Then select the corresponding metric.
Raw collected metrics and logical metrics derived through calculations or unit conversions can share the same name. Configure charts and alerts based on the metric and unit you actually use. Do not infer the unit from the _bytes suffix alone.
Unit conversions align units but do not resolve differences in the resources measured or calculation methods. Reassess thresholds against actual OB Cloud monitoring data.
If the available metrics do not meet your business needs, contact OB Cloud technical support to describe the metrics you need and how they should be defined.

Metric differences

CMS metric	Corresponding OB Cloud metric or availability	CMS definition	OB Cloud definition and differences	Usage guidance
`cpu_percent`	Same name, different definition.	Uses host CPU data when available; otherwise, uses container data.	Uses container CPU data when available; otherwise, uses host data. Aggregation methods, rate calculations, and CPU quota sources also differ.	Account for differences in data sources and calculation methods when comparing CPU usage. Reassess alert thresholds against actual OB Cloud monitoring data.
`memory_percent`	Same name, different definition.	Calculates memory usage from host free memory, cache, and buffers.	When host data is unavailable, calculates memory usage from the container working set and memory limit. The resources measured may differ from those in CMS.	Confirm whether the metric measures the host or the container before adjusting alert thresholds.
`qps`	Maps to `sql_select_count`.	Number of SELECT requests processed per second by the database.	The OB Cloud metric named `qps` measures SQL requests at the OceanBase Database Proxy (ODP) layer. The database SELECT request rate is available as `sql_select_count`.	Use `sql_select_count` and update queries, charts, and alerts accordingly. Do not substitute the OB Cloud metric named `qps`.
`qps_rt`	Maps to `sql_select_rt`.	Average response time calculated from the duration and count of SELECT requests processed by the database.	The OB Cloud metric named `qps_rt` measures the average response time of SQL requests at the ODP layer. The average response time of database SELECT requests is available as `sql_select_rt`, in milliseconds (ms).	Use `sql_select_rt` and update queries, charts, and alerts accordingly. Do not substitute the OB Cloud metric named `qps_rt`.
`ob_clog_io`	Same name, different definition.	Sum of read IOPS across all disks and write IOPS on log disks.	Measures read and write IOPS on log disks only, covering a different set of disks than CMS.	Use the OB Cloud metric to monitor log disk IOPS. Adjust charts and alert thresholds to reflect the log disks measured.
`ob_tenant_log_disk_used_bytes`	Logical metric with the same name but a different unit.	Tenant log disk usage, in bytes.	The logical metric with the same name is measured in GiB.	Display values in GiB and divide existing thresholds in bytes by 1,073,741,824.
`ob_tenant_vector_mem_hold_bytes`	Collected metric with the same name but a different unit.	Reserved vector memory, aggregated and converted to MiB. The metric name retains the `_bytes` suffix.	Raw collected values are in bytes.	Display raw collected values in bytes and multiply existing thresholds in MiB by 1,048,576.
`ob_tenant_vector_mem_used_bytes`	Collected metric with the same name but a different unit.	Used vector memory, aggregated and converted to MiB. The metric name retains the `_bytes` suffix.	Raw collected values are in bytes.	Display raw collected values in bytes and multiply existing thresholds in MiB by 1,048,576.
`ob_tenant_vector_mem_limit_bytes`	Collected metric with the same name but a different unit.	Vector memory limit, aggregated and converted to MiB. The metric name retains the `_bytes` suffix.	Raw collected values are in bytes.	When using raw collected values, multiply existing thresholds in MiB by 1,048,576. Use the same unit when comparing memory usage with the limit.
`ob_disk_free_bytes`	Same name, different definition.	Includes disks for replicas other than arbitration and log replicas.	Includes disk data for all replica types.	Account for differences in the replica types included when comparing disk space. Reassess charts and alert thresholds accordingly.
`ob_disk_total_bytes`	Same name, different definition.	Includes disks for replicas other than arbitration and log replicas.	Includes disk data for all replica types.	Account for differences in the replica types included when comparing disk space. Reassess charts and alert thresholds accordingly.
`net_throughput`	Same name, different definition.	Incoming traffic includes only network interfaces matching `eth.*`. Outgoing traffic has no equivalent restriction.	Both incoming and outgoing traffic include only network interfaces matching `eth.*`. Container network data is also supported.	Check the network interfaces and data sources included before adjusting charts and alert thresholds.
`ob_clog_io_read_time`	Same name, different definition.	Calculates the ratio of read duration to read count separately for each device.	Sums read durations and read counts before calculating their ratio. Results may differ when multiple devices are involved.	Configure charts using the average read duration calculated from the aggregated data, and reassess alert thresholds.
`ob_clog_io_time`	Same name, different definition.	Calculates the ratio of read and write duration to read and write count separately for each device.	Sums read and write durations and counts before calculating their ratio.	Configure charts using the average read and write duration calculated from the aggregated data, and reassess alert thresholds.
`ob_clog_io_write_time`	Same name, different definition.	Calculates the ratio of write duration to write count separately for each device.	Sums write durations and write counts before calculating their ratio. Results may differ when multiple devices are involved.	Configure charts using the average write duration calculated from the aggregated data, and reassess alert thresholds.
`ob_cpu_percent`	Same name, different definition.	Measures CPU usage only for tenants other than arbitration and log tenants.	Includes tenant CPU usage for all replica types.	Account for differences in the replica types included when comparing tenant CPU usage. Reassess charts and alert thresholds accordingly.
`ob_host_iops_used_percent`	Same name, different definition.	Measures IOPS utilization for devices other than `dm` devices.	Includes only OceanBase data and log disks and excludes system disk thresholds. Both the devices and thresholds included differ.	Evaluate IOPS utilization for OceanBase data and log disks.
`ob_host_real_time_iops`	Same name, different definition.	Measures IOPS for all devices other than `dm` devices.	Includes only OceanBase data and log disks, covering a different set of disks.	Adjust charts and alert thresholds to reflect OceanBase data and log disks.
`ob_host_throughput_used_percent`	Same name, different definition.	Measures throughput utilization for all devices other than `dm` devices.	Includes only OceanBase data and log disks, covering a different set of disks.	Evaluate throughput utilization for OceanBase data and log disks.
`ob_tenant_memory_percent`	Same name, different definition.	Calculates the ratio of used to total tenant context memory within the scope defined by CMS.	Calculates memory usage from tenant memory statistics. The data sources and definitions of the numerator and denominator differ.	Values may differ because of the calculation methods. When migrating alerts, reassess thresholds against actual OB Cloud monitoring data.
`system_event_internal_time_waited`	Same name, different definition.	Cumulative wait time.	Average time per wait.	Adjust queries, charts, and alerts to use the average time per wait. Do not reuse thresholds for cumulative wait time without adjustment.
`tps`	Same name, different definition.	Number of DML write operations processed per second by the database.	Number of transactions processed per second at the ODP layer. This measures a different activity from the database DML operation count.	To compare with the CMS DML write operation rate, sum `sql_insert_count`, `sql_replace_count`, `sql_update_count`, and `sql_delete_count` for the same resources and time range. Each of these four metrics measures the number of corresponding SQL statements processed per second.
`tps_rt`	Same name, different unit.	Average execution time of database DML write operations, in milliseconds (ms).	Average execution time of database DML write operations, in microseconds (μs).	Display values in μs, multiply existing thresholds in ms by 1,000, and update charts and alerts accordingly.
`transaction_rt`	Same name, different definition.	Sum of the average transaction commit time and the average rollback time.	Average processing time per transaction, including both committed and rolled-back transactions.	Use the OB Cloud metric to display the average processing time per transaction. Reassess alert thresholds configured in CMS.