Thomas Warner created HTTPCLIENT-2114:
-----------------------------------------
Summary: AsyncHttpRequestRetryExec ignores
HttpRequestRetryStrategy
Key: HTTPCLIENT-2114
URL: https://issues.apache.org/jira/browse/HTTPCLIENT-2114
Project: HttpComponents HttpClient
Issue Type: Improvement
Components: HttpClient (async)
Affects Versions: 5.0.1
Reporter: Thomas Warner
It is noted in the Javadoc that AsyncHttpRequestRetryExec does not respect
[{{HttpRequestRetryStrategy.getRetryInterval()}}|https://hc.apache.org/httpcomponents-client-5.0.x/httpclient5/apidocs/org/apache/hc/client5/http/HttpRequestRetryStrategy.html#getRetryInterval(org.apache.hc.core5.http.HttpResponse,%20int,%20org.apache.hc.core5.http.protocol.HttpContext)]
Instead, the request is retried instantly. The immediate retry happens for both
IOException retries and server responses with 429 and 503 status codes. The
Retry-After header is also ignored due to this issue. Retrying instantly is not
desirable, especially and can actively make the situation worse.
The most relevant code block is in the completed() method of the
AsyncExecCallback that is created. If the request is meant to be retried,
AsyncHttpRequestRetryExec immediately calls the internalExecute() method to
retry the request.
A naive approach would be to add to the internal State class a field for a
TimeValue representing the delay for the request. In handleResponse(), the
response from retryStrategy.retryRequest() could be stored in the state object
and then in completed(), you could call TimeValue.sleep() on that property.
However, I imagine this has the potential to lock up resources for a
significant amount of time.
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
