[ https://issues.apache.org/jira/browse/CASSANDRA-17782?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17573783#comment-17573783 ]
Doug Rohrer commented on CASSANDRA-17782: ----------------------------------------- Likely the only thing that really matters in any way is that we register the {{FailureDetector}} instance as an MBean with the default implementation (as its constructor registers itself) - I don't +think+ this is really an issue, and I don't see the failure detector actually being +touched+ by anything at runtime in tests, so it's probably fine to just call {{setDefaultFailureDetector}} and leave the patch as simple as possible. I'd appreciate feedback on that though. > `DatabaseDescriptor#clientInitialization` should set the `newFailureDetector` > (maybe to a NoOpFailureDetector?) > --------------------------------------------------------------------------------------------------------------- > > Key: CASSANDRA-17782 > URL: https://issues.apache.org/jira/browse/CASSANDRA-17782 > Project: Cassandra > Issue Type: Bug > Reporter: Doug Rohrer > Priority: Normal > Fix For: 4.1.x > > > With the introduction of a configurable FailureDetector, client users of the > {{cassandra-all}} jar now need to call not just `clientInitialization` but > also `setDefaultFailureDetector` in order to avoid an NPE when the failure > detector is constructed. This is a breaking change from previous versions > where this was not necessary. > Calls to {{DatabaseDescriptor#clientInitialization}} should set > {{newFailureDetector}} to +something+, either by calling > {{setDefaultFailureDetector}} or creating a different (NoOp?) FailureDetector. -- This message was sent by Atlassian Jira (v8.20.10#820010) --------------------------------------------------------------------- To unsubscribe, e-mail: commits-unsubscr...@cassandra.apache.org For additional commands, e-mail: commits-h...@cassandra.apache.org