4.7 KiB
| title |
|---|
| api-breaker |
Description
The plugin implements API fuse functionality to help us protect our upstream business services.
About the breaker timeout logic
the code logic automatically triggers the unhealthy state incrementation of the number of operations.
Whenever the upstream service returns a status code from the unhealthy.http_statuses configuration (e.g., 500), up to unhealthy.failures (e.g., three times) and considers the upstream service to be in an unhealthy state.
The first time unhealthy status is triggered, broken for 2 seconds.
Then, the request is forwarded to the upstream service again after 2 seconds, and if the unhealthy.http_statuses status code is returned, and the count reaches unhealthy.failures again, broken for 4 seconds.
and so on, 2, 4, 8, 16, 32, 64, ..., 256, 300. 300 is the maximum value of max_breaker_sec, allow users to specify.
In an unhealthy state, when a request is forwarded to an upstream service and the status code in the healthy.http_statuses configuration is returned (e.g., 200) that healthy.successes is reached (e.g., three times), and the upstream service is considered healthy again.
Attributes
| Name | Type | Requirement | Default | Valid | Description |
|---|---|---|---|---|---|
| break_response_code | integer | required | [200, ..., 599] | Return error code when unhealthy | |
| max_breaker_sec | integer | optional | 300 | >=3 | Maximum breaker time(seconds) |
| unhealthy.http_statuses | array[integer] | optional | {500} | [500, ..., 599] | Status codes when unhealthy |
| unhealthy.failures | integer | optional | 3 | >=1 | Number of consecutive error requests that triggered an unhealthy state |
| healthy.http_statuses | array[integer] | optional | {200} | [200, ..., 499] | Status codes when healthy |
| healthy.successes | integer | optional | 3 | >=1 | Number of consecutive normal requests that trigger health status |
How To Enable
Here's an example, enable the api-breaker plugin on the specified route.
Response 500 or 503 three times in a row to trigger a unhealthy. Response 200 once in a row to restore healthy.
curl "http://127.0.0.1:9080/apisix/admin/routes/1" -H 'X-API-KEY: edd1c9f034335f136f87ad84b625c8f1' -X PUT -d '
{
"plugins": {
"api-breaker": {
"break_response_code": 502,
"unhealthy": {
"http_statuses": [500, 503],
"failures": 3
},
"healthy": {
"http_statuses": [200],
"successes": 1
}
}
},
"upstream": {
"type": "roundrobin",
"nodes": {
"127.0.0.1:1980": 1
}
},
"uri": "/hello",
}'
Test Plugin
Then. Like the configuration above, if your upstream service returns 500. 3 times in a row. The client will receive a 502 (break_response_code) response.
$ curl -i -X POST "http://127.0.0.1:9080/hello"
HTTP/1.1 502 Bad Gateway
Content-Type: application/octet-stream
Connection: keep-alive
Server: APISIX/1.5
... ...
Disable Plugin
When you want to disable the api-breaker plugin, it is very simple, you can delete the corresponding json configuration in the plugin configuration, no need to restart the service, it will take effect immediately:
curl http://127.0.0.1:9080/apisix/admin/routes/1 -H 'X-API-KEY: edd1c9f034335f136f87ad84b625c8f1' -X PUT -d '
{
"uri": "/hello",
"upstream": {
"type": "roundrobin",
"nodes": {
"127.0.0.1:1980": 1
}
}
}'
The api-breaker plugin has been disabled now. It works for other plugins.