Wei-Chiu Chuang created HDDS-16097:
--------------------------------------
Summary: Add unit tests for
OzoneManagerRatisServer.checkRetryCache (HDDS-13621)
Key: HDDS-16097
URL: https://issues.apache.org/jira/browse/HDDS-16097
Project: Apache Ozone
Issue Type: Task
Components: Ozone Manager
Reporter: Wei-Chiu Chuang
Assignee: Wei-Chiu Chuang
Follow-up test coverage for
[HDDS-13621|https://issues.apache.org/jira/browse/HDDS-13621] (fixed in [PR
#9711|https://github.com/apache/ozone/pull/9711]).
*Problem*
When Ratis RetryCache hits a *failed* {{RaftClientReply}}
({{!reply.isSuccess()}}), the old {{checkRetryCache()}} path called
{{getOMResponse(reply)}} and could NPE because failed replies often have no
usable response message. The fix returns {{null}} so the normal submit path can
retry.
The PR was validated manually (freon ommg under JVM pauses) but has no
automated unit test.
*Scope*
Add two unit tests in {{TestOzoneManagerRatisServer}} (or a small dedicated
test class), using mocked {{RetryCache}} / {{RetryCache.Entry}} and
{{Server.getCurCall()}} setup (same pattern as {{testRetryCacheWithDownedOM}}):
# *Failed cached reply* — mock cache entry whose {{getReplyFuture()}} completes
with {{setSuccess(false)}} / {{Message.EMPTY}} (e.g. {{NotLeaderException}}).
Assert {{checkRetryCache()}} returns {{null}} and does not NPE.
# *Successful cached reply* — mock cache entry with {{setSuccess(true)}} and a
serialized {{OMResponse}} in the reply message. Assert {{checkRetryCache()}}
returns the expected {{OMResponse}} (regression guard).
*References*
* Parent fix: HDDS-13621 / PR #9711
* Related idempotent retry work: HDDS-11558
* Existing integration coverage (success path only):
{{TestOzoneManagerHAWithAllRunning#testOMRetryCache}},
{{TestOzoneManagerHAWithStoppedNodes#testRetryCacheWithDownedOM}}
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
