Hi, I agree with Yufei: 1. I would announce EclipseLink will be removed in 1.3 2. We do remove it in the 1.3 release 3. I don't think we need any tool: moving from EclipseLink to JDBC should be smooth and with minimal effort. For one shot effort, not sure it's worth to spend time on "migration tool".
Regards JB On Thu, Sep 18, 2025 at 1:41 AM Yufei Gu <flyrain...@gmail.com> wrote: > > +1 on removing it. Given Polaris’ monthly release cadence, it seems fine to > wait two (remove in 1.3) or three (remove in 1.4) more releases. > > https://github.com/apache/polaris-tools/tree/main/polaris-synchronizer can > migrate principals, but doesn't support policies. > > I’m not sure it’s worth building another type of migration tool for this > use case, we might be better off improving the existing ones. > > > Yufei > > > On Wed, Sep 17, 2025 at 1:10 PM Adam Christian < > adam.christian.softw...@gmail.com> wrote: > > > You are right, Russell. We should make a clear migration path, so our > > EclipseLink users are able to easily transition off on EclipseLink. I know > > that this has come up before [1]. Let me investigate a few options on what > > guidance we can give or what tooling we can produce. > > > > [1] https://github.com/apache/polaris/issues/1875 > > > > Cheers, > > > > Adam > > > > On Wed, Sep 17, 2025 at 3:49 PM Dmitri Bourlatchkov <di...@apache.org> > > wrote: > > > > > We have two migration tools: > > > > > > * > > > > > https://github.com/apache/polaris-tools/tree/main/iceberg-catalog-migrator > > > > > > * https://github.com/apache/polaris-tools/tree/main/polaris-synchronizer > > > > > > I'm pretty confident that iceberg-catalog-migrator works well, but it can > > > only migrate tables, not principals. > > > > > > I never personally used polaris-synchronizer, still it's supposed to > > > migrate all Polaris data, including principals. > > > > > > Cheers, > > > Dmitri. > > > > > > On Wed, Sep 17, 2025 at 3:13 PM Russell Spitzer < > > russell.spit...@gmail.com > > > > > > > wrote: > > > > > > > +1 I think removing EclipseLink should happen soon now that we have 2 > > > > releases with it deprecated. I have > > > > looked too deeply into this but do we have a migration plan for users > > > > already on EclipseLink to get over to the > > > > JDBC Impl? > > > > > > > > On Wed, Sep 17, 2025 at 12:53 PM Dmitri Bourlatchkov <di...@apache.org > > > > > > > wrote: > > > > > > > > > Thanks for bringing this issue up, Adam! > > > > > > > > > > I support removing EclipseLink code immediately. > > > > > > > > > > My rationale: > > > > > > > > > > * Due to EclipseLink deprecation, non-trivial new features are not > > > > > necessarily implemented there [1] > > > > > > > > > > * Any new bugs reported for EclipseLink are not likely to get > > attention > > > > > because this backend is in decline. > > > > > > > > > > * Users had better migrate to a supported backend earlier. If > > migration > > > > is > > > > > deferred, it will likely mean that any issues related to migration > > will > > > > > take even longer to be found. > > > > > > > > > > * Polaris 1.1.0 still has EclipseLink, which offers users a supported > > > > > version where critical issues could still be fixed, if they are > > found. > > > > > > > > > > * Having EclipseLink in the codebase adds overhead for new features > > > that > > > > > touch Persistence. > > > > > > > > > > [1] > > > > > > > > > > > > > > > > > > > https://github.com/apache/polaris/pull/2197/files#diff-59a870c7af1578200236f22d35fd2eb75dc2a1e73e51218464eb7ba089217da7R759 > > > > > > > > > > Cheers, > > > > > Dmitri. > > > > > > > > > > On Wed, Sep 17, 2025 at 1:27 PM Adam Christian < > > > > > adam.christian.softw...@gmail.com> wrote: > > > > > > > > > > > Howdy Polaris Community! > > > > > > > > > > > > I was going through our open bugs and I noticed that there are > > > around 5 > > > > > to > > > > > > 10 bugs related to EclipseLink persistence. I was wondering when we > > > > > > believe a good time to remove EclipseLink would be. > > > > > > > > > > > > Personally, I think we could probably start doing it now since it's > > > > been > > > > > > deprecated since 1.0.0 and we have a clear alternative. I believe > > > there > > > > > are > > > > > > several pros for our users such as streamlined documentation and > > > > benefits > > > > > > to the contributors such as less issues, dependencies, and modules. > > > > > > > > > > > > How do y'all feel about this? > > > > > > > > > > > > If we are aligned, I can create the issue and start working on it. > > > > > > > > > > > > Cheers, > > > > > > > > > > > > Adam > > > > > > > > > > > > > > > > > > > >
