[
https://issues.apache.org/jira/browse/SCB-1974?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17157822#comment-17157822
]
Likun Liu commented on SCB-1974:
--------------------------------
We'd treat all 500 errors the same way as 503 in terms of isolation
requirement, and also include the server timeout error (408) for isolation,
which indicates the server is under stress.
> LoadBalance isolation feature should trigger on provider errors
> ---------------------------------------------------------------
>
> Key: SCB-1974
> URL: https://issues.apache.org/jira/browse/SCB-1974
> Project: Apache ServiceComb
> Issue Type: Bug
> Reporter: Likun Liu
> Priority: Major
>
> In LoadBalance module, there is isolation feature which triggers on the error
> rate from providers. However, the way to check if the request is failed or
> not seems wrong:
> [https://github.com/apache/servicecomb-java-chassis/blob/master/handlers/handler-loadbalance/src/main/java/org/apache/servicecomb/loadbalance/LoadbalanceHandler.java#L431-L432]
> {code:java}
> // code placeholder
> return e.getStatusCode() ==
> ExceptionFactory.CONSUMER_INNER_STATUS_CODE || e.getStatusCode()
> == 503;
> {code}
> We'd also include `ExceptionFactory.PRODUCER_INNER_STATUS_CODE`, which
> implies the error from providers.
> I'm not sure why `ExceptionFactory.CONSUMER_INNER_STATUS_CODE` is used in the
> check, it doesn't help to isolate on providers since the error is from
> consumer side. Maybe there are other reasons (maybe for the stats?) for this
> error code is used.
>
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
