Use TopRegion Analysis to diagnose hot shards - Lindorm

When traffic is unevenly distributed across a Lindorm cluster, some regions absorb a disproportionate share of reads, writes, or data volume. This causes elevated response times and degrades cluster performance. Use the TopRegion Analysis feature to identify these hot regions in real time or at a specific historical point in time, so you can take action before performance is affected.

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.

Analyze hot regions in real time

Use the TopRegion Realtime Analysis section to query the current hot regions.

In the left-side navigation pane, choose Diagnose Tools > topRegion Analysis.

In the TopRegion Realtime Analysis section, set the filter conditions. The following sort dimensions are available:

Filter	Description
idc	The data center (IDC) to query.
Group	The server group to query.
Server	The specific server to query.
Sort dimension	The metric by which results are ranked. See the sort dimensions table below.
Table name	The table to filter results by.
Top	The number of top results to display.

Dimension	Unit	Description
Total requests	Requests/second	The number of requests sent per second to access a shard.
Read requests	Requests/second	The number of requests sent per second to read a shard.
Write requests	Requests/second	The number of requests sent per second to write a shard.
Shard size	MB	The size of a shard.
File nums	Files	The total numbers of files in each shard.
Read response time	Millisecond	The average response time for requests sent to read a shard.
Write response time	Millisecond	The average response time for requests sent to write a shard.
Memstore Size	MB	The amount of data in the memory.
Total request data	KB/second	The amount of data generated by all requests sent per second to access a shard.
Read data	KB/second	The amount of data generated by all read requests sent per second to read a shard.
Write data	KB/second	The amount of data generated by all write requests sent per second to write a shard.

Note

Turn on Table aggregation to group results by table. Turn on Ascending to sort results in ascending order by the selected dimension.

Click Query. The following example shows a query with Sort dimension set to Read requests, Top set to 5, and both Ascending and Table aggregation turned on. The results display the five shards with the most read requests, aggregated by table and sorted in ascending order.

Analyze hot regions from historical snapshots

Use the TopRegion History snapshot section to query hot regions at a specific point in the past.

Note

Historical snapshots are recorded at a 30-second interval.

In the TopRegion History snapshot section, set the filter conditions.
Filter Description
idc The data center (IDC) to query.
Group The server group to query.
History snapshot The historical point in time at which to query hot regions.
Click Query. The results are organized into two sections:
- Request Top: The top 10 tables with the most requests, along with the shards where they are stored.
- Vol Top: The top 10 tables with the largest amount of request data, along with the shards where they are stored.
(Optional) Click the icon next to a table to view the details of the shard where it is stored.