In order to keep the momentum and avoid getting stuck with Druid for the next 5 years I logged HIVE-29745 [1] and raised a PR [2]. I am hoping that by next week master will have 30K lines of code less so please take a look and share your thoughts.
Best, Stamatis [1] https://issues.apache.org/jira/browse/HIVE-29745 [2] https://github.com/apache/hive/pull/6615 On Fri, Jul 17, 2026 at 3:21 PM Simhadri G <[email protected]> wrote: > +1 for removal. > > There was already a thread to discuss removing Druid 4 years ago! [1] > > There was no response from users using druid back then either. > > [1] https://lists.apache.org/thread/1swonwzts4rztz0vdjhllbzmotsdx8ro > > > Thanks! > Simhadri G > > On Fri, Jul 17, 2026 at 5:44 PM Stamatis Zampetakis <[email protected]> > wrote: > >> +1 for immediate removal without deprecation. >> >> I checked the state of the Druid module today and it is completely >> broken. The Druid 0.17.1 version is very old and is incompatible with >> many of our core dependencies (calcite, avatica, jackson, derby, JDK). To >> be more precise the module has been broken since at least 2021 where we >> bumped the jackson version as part of HIVE-24816. At the same time the main >> end-to-end tests for the module >> (namely TestMiniDruidCliDriver, TestMiniDruidKafkaCliDriver) have been >> permanently disabled. >> >> To get a sense of how badly broken it is I tried running >> TestMiniDruidCliDriver >> with druidmini_test1.q and unsurprisingly failed since Druid's broker, >> coordinator, historical, cannot come up. I spent a bit of time trying to >> restore some dependencies (jackson, calcite, avatica, etc.) to be >> compatible with the old version but eventually hit JDK incompatibilities >> and abandoned the effort. >> >> For the past 5 years, nobody cared or complained about the broken Druid >> module so I think it's safe to infer that nobody is using it anymore. >> >> Best, >> Stamatis >> >> On Thu, Jul 16, 2026 at 8:04 PM Ayush Saxena <[email protected]> wrote: >> >>> +1 >>> >>> If we conclude that nobody, or almost nobody, is using it, and keeping >>> it is creating unnecessary resistance to other important initiatives, I’m >>> fine with removing it in 4.3 as well—although that would be quite an >>> aggressive move >>> >>> -Ayush >>> >>> On 16 Jul 2026, at 10:25 PM, Attila Turoczy <[email protected]> >>> wrote: >>> >>> >>> Totally agree with Sai's points. I also agree to depricate this module. >>> >>> -Attila >>> >>> On Thu, Jul 16, 2026 at 6:44 PM Sai Hemanth Gantasala < >>> [email protected]> wrote: >>> >>>> Hello team, >>>> >>>> I support deprecating druid-handler and targeting removal in a future >>>> major release. >>>> >>>> I worked on HIVE-29707 (upgrade Druid 0.17.1 --> 26.0.0) to address CVE >>>> exposure from the 2020-era dependency. The handler itself can be made to >>>> compile, but keeping it current is costly and the blast radius is wide: >>>> 1) Large API drift across storage handler, indexing, serde, and Kafka >>>> supervisor classes >>>> 2) hive-it-druid (~200 MB shaded jar) is pulled into unrelated qfile >>>> drivers (e.g. TestMiniLlapCliDriver), causing collateral test failures >>>> 3) Ongoing shading conflicts with Hive's stack (Calcite/Jackson/Guava), >>>> consistent with the CALCITE-6393 issues mentioned above >>>> Combined with minimal recent development and no clear maintainer/user >>>> base, I'd deprecate in docs/build now and remove in the next major. I >>>> wouldn't invest in a full upgrade unless someone commits to long-term >>>> ownership. >>>> >>>> Happy to share HIVE-29707 details if useful for a formal deprecation >>>> proposal. >>>> >>>> On 2026/07/15 11:38:13 Ruben Q L wrote: >>>> > Hello! >>>> > >>>> > In the context of a WIP Calcite upgrade [1] I ran into shading issues >>>> in >>>> > druid-handler (an ASM failure when relocating Calcite's >>>> SqlFunctions.class, >>>> > tracked upstream in CALCITE-6393). I worked around it by dropping the >>>> > Calcite relocation in druid-handler/pom.xml, but it prompted me to >>>> re-open >>>> > an earlier discussion [2] about whether we should still be carrying >>>> this >>>> > module at all. >>>> > >>>> > Nothing seems to have changed since it was last raised: we're still on >>>> > Druid 0.17.1 [3] (released in 2020), which has a number of unpatched >>>> CVEs >>>> > [4], and there's been essentially no meaningful development in the >>>> module >>>> > [5]. >>>> > >>>> > So the same questions remain open: >>>> > - Are there any active users of druid-handler? >>>> > - If there are users but no maintainers, should we deprecate the >>>> module >>>> > (docs + build) and target removal in a future major release? >>>> > - If there are neither, shall we drop it now? >>>> > >>>> > Best, >>>> > Ruben >>>> > >>>> > [1] https://issues.apache.org/jira/browse/HIVE-29641 >>>> > [2] https://lists.apache.org/thread/1swonwzts4rztz0vdjhllbzmotsdx8ro >>>> > [3] >>>> > >>>> https://github.com/apache/hive/blob/c4791205d84201b351a8353e9eacadbc3c972c3e/pom.xml#L137 >>>> > [4] https://mvnrepository.com/artifact/org.apache.druid/druid/0.17.1 >>>> > [5] >>>> > >>>> https://github.com/search?q=repo%3Aapache%2Fhive+druid&type=commits&s=committer-date&o=desc >>>> > >>>> >>>
