Jose Luis López created HADOOP-20048:
----------------------------------------
Summary: Fix race in TestHttpServer#testHttpServer2Metrics
Key: HADOOP-20048
URL: https://issues.apache.org/jira/browse/HADOOP-20048
Project: Hadoop Common
Issue Type: Bug
Components: test
Reporter: Jose Luis López
{{TestHttpServer#testHttpServer2Metrics}} is intermittently flaky. It was
observed failing once and passing on retry (surefire {{flakyFailure}}) with:
{noformat}
java.lang.AssertionError:
Expecting:
<104>
to be greater than:
<104>
{noformat}
The test reads {{HttpServer2Metrics#responses2xx()}} immediately after
{{conn.getResponseCode()}} returns 200. That counter comes from Jetty's
{{StatisticsHandler}}, which records a response when the request completes on
the server side. This can happen after the client has already read the status
line, so the second read can still return the old value.
Proposed fix (test-only):
* read and close the response body, then {{disconnect()}}, so the exchange
completes;
* wait for the counter with {{GenericTestUtils.waitFor(() ->
metrics.responses2xx() > before, 10, 10000, ...)}} instead of reading it once.
The test still checks the same thing: the 2xx counter increases after a
successful request. With the fix, the test passed 50/50 runs in separate JVMs
with {{-Dsurefire.rerunFailingTestsCount=0}}, and checkstyle reports no new
warnings.
--
This message was sent by Atlassian Jira
(v10.3.26#10030026)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
