Tair (Redis OSS-compatible) instances are located at the data layer, close to application services. They often handle a high volume of data access requests and consume network bandwidth. The maximum bandwidth varies by instance type. Exceeding this limit affects the access performance of your applications.
Step 1: Query traffic usage
You can query the traffic usage of the instance for a specific time period. If you know when the traffic spike occurred, you can skip this step and proceed to Step 2: Quick troubleshooting.
In this example, the inbound and outbound traffic rapidly increases and remains at 100%, as shown in the following figure:
Typically, if the average traffic usage consistently remains at 80% or higher, you should investigate the issue. This may indicate that the bandwidth is insufficient.
The metrics to follow are inbound traffic utilization (Intranet In Ratio) and outbound traffic utilization (Intranet Out Ratio).
Figure 1. Traffic usage example
Step 2: Quick troubleshooting
Multiple issues can cause a traffic spike on an instance. Check the following items one by one to identify the cause.
Before you begin troubleshooting, you can temporarily adjust the instance bandwidth in an emergency. This helps reduce the impact on your services and provides a larger time window to investigate the problem.
Large keys and hot keys
First, use the Top Key Statistics feature to check for large keys or hot keys. If any are found, the feature displays specific information about the keys in the console.
Impact: Large keys cause traffic spikes, while hot keys cause a sustained increase in traffic.
If large keys exist, split them based on business logic, such as user ID or time range. Reduce access to them, or delete unnecessary large keys. For more information, see Large keys and hot keys.
If hot keys exist, you can also split them based on your business logic, or use the Proxy Query Cache feature to cache the hot keys.