How to Handle and Understand the Reason for tikvclient Backoff Seconds Count Alert

Translated
1

Note:
This topic has been translated from a Chinese forum by GPT and might contain errors.

Original topic: tikvclient backoff seconds count的报警该如何处理是啥原因

| username: TiDBer_oHSwKxOH

How should the alarm for tikvclient backoff seconds count be handled and what is the cause?

2
| username: 像风一样的男子 | Original post link

It is likely caused by retries due to region or leader scheduling. The scheduling and splitting of regions, as well as leader scheduling, can trigger this alert. If the latency requirement is not high, you can modify the threshold to avoid triggering the alert.

3
| username: WalterWj | Original post link

What are the exception types in backoff?

4
| username: 大飞哥online | Original post link

You can check the network issues between environments, capture packets or ping large packets to see if there is any loss. If the alerts are frequent, you can appropriately adjust the threshold.

5
| username: Kongdom | Original post link

You can refer to this post:

image
image751×310 7.49 KB

6
| username: tidb菜鸟一只 | Original post link

Backoff is unavoidable, and the busier the system, the more it happens. What I mean is, if it keeps reporting, either adjust the threshold, change the development logic, or directly upgrade the system configuration.

7
| username: redgame | Original post link

High network latency is a major drawback.