请求重试策略介绍
Load Balancing is capable of redirecting a request to an alternative origin for retrial when a request to an initially designated server fails, in accordance with the retry policy. This reduces business request failures caused by network issues or origin fails.
Note:
Actual business requests may fail due to the following reasons:
1. Origin Failure and Have Not Actively Probe to Disable: After health check policies are configured, active probe is conducted periodically. Traffic is directed based on the results of the previous probe until new results are available. If an origin becomes unhealthy between two probes, business traffic might still be directed to the unhealthy origin. This leads to business request failure.
2. Network Jitter: The origin is healthy, but network issues occur during the access. This leads to business request failure.
Note:
Request failures include origin connection establishment failures and origin response reception failures.
For the situations mentioned above, EdgeOne provides the following two fallback request retry policies:
Policy 1: When a real business request fails to access a certain origin, it directly retries with another origin within the next lower priority origin group. This is suitable for scenarios where the performance of both higher and lower priority origin groups is similar.
Policy 2: When a real business request fails to access a certain origin, it directly retries with another origin within the same priority origin group. This is suitable for scenarios where the performance of the higher priority origin group is significantly better than that of the lower priority origin group.
Note：
POST requests do not support origin-pull retry.