Hi all, I think we should do a small 0.16.1 fix release to address two rolling upgrade issues, add improvements to the startup scripts that would help avoid some very confusing failures for new users, and address the Wikipedia dataset licensing questions that were raised during the 0.15.1 voting process. I'll volunteer to manage the release if we agree on it.
-------------------- There are two known issues with rolling upgrades that affect 0.16.0, fixed by the following PRs: - https://github.com/apache/incubator-druid/pull/8864 add sequenceName and currentCheckPoint for backwards compatibility #8864 - https://github.com/apache/incubator-druid/pull/8905 Retrying with a backward compatible task type on unknown task type error in parallel indexing #8905 -------------------- There have been quite a few people hitting issues when they try to run Druid under a Java version later than 8. While our docs mention that only Java 8 is supported, the failure mode when that condition is not met occurs pretty late: users try ingesting some data, and stuff works until they see that historicals are not loading their segments, with no immediate indication that the error is occurring because of Java version incompatibilities. The following links from the ASF Slack channel show examples of users encountering this confusing failure: - https://the-asf.slack.com/archives/CJ8D1JTB8/p1569948481021000 - https://the-asf.slack.com/archives/CJ8D1JTB8/p1572936011252000?thread_ts=1572845319.226400&cid=CJ8D1JTB8 - https://the-asf.slack.com/archives/CJ8D1JTB8/p1572561066155100 The following patch adds a Java 8 version check to the bundled startup scripts, so that users are given a clear error reason promptly: https://github.com/apache/incubator-druid/pull/8794 Startup scripts: verify Java 8 (exactly), improve port/java verification messages. A follow-on patch to the above also addresses port conflicts that can result in confusing behavior for people getting started with Druid (the failure mode here is that a user has an existing process that binds only on localhost to a port that Druid uses, so traffic destined for localhost with that port hits that non-Druid process but traffic destined for other IP addresses on that machine will go to Druid): https://github.com/apache/incubator-druid/pull/8942 Improve verify-default-ports to check both INADDR_ANY and 127.0.0.1. -------------------- The licensing questions around the bundled Wikipedia dataset that were raised in the IPMC vote for 0.15.1 have been resolved ( https://issues.apache.org/jira/browse/LEGAL-480), so we should also update our licensing-related documentation soon. Thanks, Jon
