Hi, I'm in favor of this. In fact, I had already done the work to bump us all the way to compiling with JDK 25, but this has not yet been merged due to the Druid v38 release requirement of supporting both 21 and 25. We already are running CI with JDK 25 on 21-compiled bytecode.
On Mon, Jul 6, 2026 at 12:34 PM Karan Kumar <[email protected]> wrote: > Hey Folks > I’m collating the various threads we have around JDK version support into > one mega thread: > 1. Dropping support of jdk 17 thread > <https://lists.apache.org/thread/q7mgcd11owch7k6vx8cvb32n7t2d4hfv>. > 2. Dropping support of jdk 11 thread > <https://lists.apache.org/thread/gy213gq134bxhdkdprm6fj84c860xdg0>. > > I think, as a community, we have aligned that JDK 25 is the future-facing > path. Moving to it would let us start using newer language/runtime features > like virtual threads and SIMD/vector improvements, which should help > improve Druid performance in general. > > The main blocker seems to be support for the ORC and parquet readers and > with Hadoop 3.5 now released, we could try upgrading to it and see whether > that resolves the compatibility gaps. > > I wanted to get a sense of how folks feel about making Druid 39 the release > where we fully migrate to JDK 25 as the minimum supported Java runtime > version. >
