Big Call Request Detection
A large query request is a Remote Procedure Call (RPC) request that scans more than 100 data blocks, caused by improperly specified query conditions. These requests consume significant CPU and I/O resources and can degrade system throughput.
Use the Big Call Request Detection tool in the cluster management system to identify large query requests at the instance or node level, so you can investigate and resolve performance issues promptly.
Prerequisites
Before you begin, ensure that you have:
Logged on to the cluster management system of your Lindorm instance. For more information, see Log on to the cluster management system
Detect large query requests
In the left-side navigation pane, choose Diagnose Tools > Big Call Request Detection.
Set the query conditions, including idc, Group, and Server.
Click Query.
Interpret the results
The query results include the following parameters.
| Parameter | Description |
|---|---|
| Request detail | Detailed information about the large query request, including the time it was initiated and the request content. |
| Sampling time | The time when the large query request was initiated. |
If no results are returned, no large query requests exist on the selected node.
TopRegion analysis
This topic describes how to view hot region analysis results in the cluster management system.
Prerequisites
You are logged on to the cluster management system of your Lindorm instance. For more information, see Log on to the cluster management system.
Background information
In database operations, a high number of files can increase request latency. At the same time, uneven traffic distribution can overload specific nodes and create hotspots, which can lead to system-wide performance degradation and disrupt normal operations. The cluster management system provides a traffic diagnosis feature to help you identify hotspots and locate hot regions.
TopRegion real-time analysis
You can perform a real-time analysis to query for hot regions.
In the left-side navigation pane of the cluster management system, choose .
Set the filter conditions, including IDC, Group, Node, sort dimension, and Table Name.
The following table describes the sort dimensions.
Dimension
Unit
Description
Total requests
Requests/second
The number of requests per second for each region.
Read requests
Requests/second
The number of read requests per second for each region.
Write requests
Requests/second
The number of write requests per second for each region.
region size
MB
The size of the region.
number of files
Count
The total number of files in each region.
read response time
ms
The average response time for read operations on each region.
write response time
ms
The average response time for write operations on each region.
Memstore Size
MB
The size of data in the Memstore.
Total request data volume
KB/s
The total data volume of requests per second for each region.
Read data volume
KB/s
The data volume of read requests per second for each region.
Write data volume
KB/s
The data volume of write requests per second for each region.
NoteIn the Top text box, enter the number of records to display in the query results.
Enable table aggregation to display aggregated data.
Enable Ascending order to sort the results in ascending order.
Click Query.
For example, consider a query where the sort dimension is set to
Read requests, Top is set to5, and both the Ascending order and table aggregation switches are enabled. This query first identifies the five regions with the highest number of read requests per second. Then, it groups the regions that belong to the same table. Finally, it sorts the aggregated results in ascending order by the number of read requests.
TopRegion historical snapshot
You can use historical snapshots to query for hot regions at a specific point in the past.
Select a IDC, Group, and Historical Snapshots.
NoteA Historical Snapshots represents a specific point in time. Snapshots are captured at 30-second intervals.
Click Query.
The query results are divided into the following two sections:
Top Requests: The tables with the highest total request count in the specified time period and their corresponding regions. By default, the top 10 tables are displayed.
Top Data Volume: The tables with the largest total data volume in the specified time period and their corresponding regions. By default, the top 10 tables are displayed.
(Optional) Click the
icon next to a table name to view the details of the table's regions.
Client Trace
When a Lindorm table receives unexpected write traffic, a namespace is accessed by an unknown client, or you need to audit which applications are connecting to your instance, use Client Trace to identify the source.
Client Trace provides two tracing dimensions:
User dimension: shows the last request time and client details for a specific database user. Use this when you know which user account to investigate.
Node, namespace, or table dimension: shows the IP addresses of clients performing read and write operations on a specific node group, namespace, or table. Use this when you need to trace access by cluster location rather than by user account.
Prerequisites
Before you begin, ensure that you have:
Logged on to the cluster management system of your Lindorm instance. For more information, see Log on to the cluster management system.
Trace clients by user
Use the user dimension when you want to see which clients a specific database user has connected from, and when those connections last occurred.
In the left-side navigation pane of the cluster management system, choose .
In the User dimension client trace section, select the user you want to trace from the user drop-down list.
Click Trace.
The results include the following fields:
| Field | Description |
|---|---|
| Username | The database user that initiated the request. |
| Client addr | The IP address of the client that initiated the request. |
| Client version | The version of the client used to initiate the request. |
| Last contact time | The timestamp of the most recent request from this client. |
Trace clients by node, namespace, or table
Use this dimension when you need to identify which client IP addresses are sending requests to a specific node group, namespace, or table, regardless of which user account initiated them.
In the Node, namespace, table dimension client trace section, configure the filter conditions: idc, Group, and node.
(Optional) Enter a namespace name in the namespace field and a table name in the table field to narrow the results.
Click Trace.
The results show the IP addresses of all clients that sent requests matching your filter conditions.