Il Lun 24 Apr 2023, 12:35 PengHui Li <peng...@apache.org> ha scritto:
> Can we cherry-pick fixes for the new features that were introduced to > 3.0.0? > We do lots of chaos testing, stress testing for new delayed messages, and > load balancer > before the 3.0.0 release. Is it reasonable to have the fixes without > releasing buggy > features? > I 100% agree But RM but be aware and they should do the cherry picks Enrico > Or the testing should be completed before the code freeze, otherwise, push > to the next feature release. > > Thanks, > Penghui > > On Mon, Apr 24, 2023 at 6:08 PM Christophe Bornet <bornet.ch...@gmail.com> > wrote: > > > Hi Zike, > > > > I just reverted the commits of PIP-195 (minutes before your mail 😄) > > except for > > > https://github.com/apache/pulsar/commit/e248e14473142765db1df324c20a081a2422980e > > Maybe we should keep that one since it will be a breaking change if we > > do it later. > > > > Le lun. 24 avr. 2023 à 11:52, Zike Yang <z...@apache.org> a écrit : > > > > > > Hi, all > > > > > > Based on the above discussion, I will revert all these commits. They > > > are all related to the improvement of the PIP-195: > > > * > > > https://github.com/apache/pulsar/commit/e248e14473142765db1df324c20a081a2422980e > > > * > > > https://github.com/apache/pulsar/commit/e1d63990644700bf61b3d7af1ef6d4d62145c2bb > > > * > > > https://github.com/apache/pulsar/commit/ff59240165c73a9c3a3dcca20702ab44b0b18d33 > > > * > > > https://github.com/apache/pulsar/commit/49480ea558e647169e8df01bfd2e871a5386e19e > > > * > > > https://github.com/apache/pulsar/commit/1d1a3ef864a65c995ceda4b7875ed934c2574298 > > > > > > This commit has been reverted: > > > > > > https://github.com/apache/pulsar/commit/3da39b2e1553cecbc6d6b85e8bc7844f611d5637 > > > > > > And we keep these two commits for Pulsar 3.0.0: > > > * > > > https://github.com/apache/pulsar/commit/0df741259505b9189147d72c6f013b2c18f0436c > > > * > > > https://github.com/apache/pulsar/commit/d05871213adc351d4c718c2a6fb0909b01d279ff > > > > > > @Nicolò Could you provide more context for this commit: > > > > > > https://github.com/apache/pulsar/commit/e27abe9e128fb71b65ffe06417574c9a7f3facbd > > ? > > > Do we need to include it in Pulsar 3.0.0? > > > > > > Please let me know if you have any questions. > > > > > > Thanks, > > > Zike Yang > > > Zike Yang > > > > > > On Mon, Apr 24, 2023 at 5:23 PM Zike Yang <z...@apache.org> wrote: > > > > > > > > Hi, Michael > > > > > > > > Thanks for your explanation. I'm +1 for keeping these commits. > > > > > > > > Thanks, > > > > Zike Yang > > > > > > > > On Sat, Apr 22, 2023 at 12:47 AM Michael Marshall < > > mmarsh...@apache.org> wrote: > > > > > > > > > > > So to respect the code freeze, I propose to revert all those > > commits except: > > > > > > https://github.com/apache/pulsar/pull/19849 > > > > > > https://github.com/apache/pulsar/pull/20086 > > > > > > > > > > Can you enumerate which commits you want to revert? 19849 was not > > > > > cherry picked, so it is not relevant to this discussion. > > > > > > > > > > I feel strongly that we need to keep the following commits: > > > > > > > > > > > > > https://github.com/apache/pulsar/commit/370f6d7af78fa7bd8f92f8116fac4e3cf32adecb > > > > > > > > > > Justification: I broke the AuthenticationProviderList with my > changes > > > > > from PIP 97. This commit fixes that regression from 2.11. > > > > > > > > > > > > > https://github.com/apache/pulsar/commit/b839536cb05a0e1c737d9fb87edae1e054fd301b > > > > > > > > > > Justification: this fixes a broken error log to help users quickly > > > > > identify a Kubernetes bug that users of the OIDC plugin will > > > > > definitely encounter. I spent at least 4 hours earlier this week > > > > > tracking down an issue that would have been obvious had the error > log > > > > > been correct. The change is very small and easy to understand, and > > > > > will save users a lot of time. > > > > > > > > > > > > > https://github.com/apache/pulsar/commit/0df741259505b9189147d72c6f013b2c18f0436c > > > > > > > > > > Justification: this is a one line bug fix that is necessary for > OIDC > > > > > to work in function pods. It is well understood and covered by a > > test. > > > > > > > > > > Thanks, > > > > > Michael > > > > > > > > > > > > > > > On Fri, Apr 21, 2023 at 5:27 AM Zike Yang <z...@apache.org> wrote: > > > > > > > > > > > > > I understand that the optimization PRs are nice but if we start > > to > > > > > > > accept them, why not accepting all the others and the code > freeze > > > > > > > becomes pointless. > > > > > > > We can do a 3.0.1 shortly after to include all those > > enhancements. > > > > > > > > > > > > Totally +1 for this. > > > > > > > > > > > > > The mailing list is too "slow" for this kind of stuff. For more > > complex > > > > > > > discussions, there's always the possibility to start a thread > > here. > > > > > > > > > > > > I still recommend discussing the cherry-pick on the mailing list. > > > > > > For the LTS release, we need to be very careful when > cherry-picking > > > > > > the commit. All cherry-picked commits that enter the LTS version > > > > > > should be noticed for everyone. They should all be considered as > > > > > > important changes. Another benefit of discussing it in the > mailing > > > > > > list is that we can keep the context of the cherry-pick > > permanently. > > > > > > > > > > > > As long as we have a consensus, then any committers can > > cherry-pick the commit. > > > > > > > > > > > > Thanks, > > > > > > Zike Yang > > > > > > > > > > > > On Fri, Apr 21, 2023 at 6:08 PM Cong Zhao <zhaoc...@apache.org> > > wrote: > > > > > > > > > > > > > > I agree with this, other PRs can move to 3.0.1. > > > > > > > > > > > > > > Thanks, > > > > > > > Cong Zhao > > > > > > > > > > > > > > On 2023/04/21 09:43:20 Christophe Bornet wrote: > > > > > > > > So to respect the code freeze, I propose to revert all those > > commits except: > > > > > > > > https://github.com/apache/pulsar/pull/19849 > > > > > > > > https://github.com/apache/pulsar/pull/20086 > > > > > > > > > > > > > > > > I understand that the optimization PRs are nice but if we > > start to > > > > > > > > accept them, why not accepting all the others and the code > > freeze > > > > > > > > becomes pointless. > > > > > > > > We can do a 3.0.1 shortly after to include all those > > enhancements. > > > > > > > > > > > > > > > > Do we have an agreement on this ? > > > > > > > > > > > > > > > > Christophe > > > > > > > > > > > > > > > > Le ven. 21 avr. 2023 à 11:36, Cong Zhao <zhaoc...@apache.org > > > > a écrit : > > > > > > > > > > > > > > > > > > I strongly agree that these cherry-pick should be notified > > to the release managers to avoid unintended consequences. > > > > > > > > > > > > > > > > > > PR https://github.com/apache/pulsar/pull/20086 is to solve > > a serious bug it will lead to the pulsar-io block, so it is very > necessary > > to cherry-pick into 3.0. The rest of the PRs are optimized for large > > amounts of data, we also can discuss whether they are necessary to > > cherry-pick into 3.0. > > > > > > > > > > > > > > > > > > Thanks, > > > > > > > > > Cong Zhao > > > > > > > > > > > > > > > > > > On 2023/04/21 07:26:43 Nicolò Boschi wrote: > > > > > > > > > > I believe that currently there's no clear process in > place > > in order to > > > > > > > > > > decide whether a commit should be cherry-picked into the > > frozen branch. > > > > > > > > > > I know that we have a slack channel for the release > > coordination > > > > > > > > > > #pulsar-release-3_0, which is open to everyone. > > > > > > > > > > > > > > > > > > > > One solution would be to let the release managers > > cherry-pick the commits - > > > > > > > > > > they are the contributors more focused on the stability > of > > the release > > > > > > > > > > branch and they might be able to discuss if it's > > worthwhile. > > > > > > > > > > Their opinion is not more important than others, but they > > need to know it > > > > > > > > > > anyway and they could spot incompatibilities and avoid > > unintended > > > > > > > > > > consequences. > > > > > > > > > > The mailing list is too "slow" for this kind of stuff. > For > > more complex > > > > > > > > > > discussions, there's always the possibility to start a > > thread here. > > > > > > > > > > > > > > > > > > > > When a committer would like to cherry-pick a commit, > > instead of directly > > > > > > > > > > going for it without any discussion, they can ask in the > > release slack > > > > > > > > > > channel. > > > > > > > > > > The release managers will eventually cherry-pick it. > > > > > > > > > > If there's no consensus then the discussion is moved to > > the mailing list. I > > > > > > > > > > believe this wouldn't happen often, considering that > > currently we rely on > > > > > > > > > > the common sense of committers. > > > > > > > > > > > > > > > > > > > > What do you think? > > > > > > > > > > Nicolò Boschi > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > Il giorno ven 21 apr 2023 alle ore 07:58 Cong Zhao < > > zhaoc...@apache.org> ha > > > > > > > > > > scritto: > > > > > > > > > > > > > > > > > > > > > Hi Zike, > > > > > > > > > > > > > > > > > > > > > > I'm sorry for cherry-picking the new delayed message > PRs > > to branch-3.0. > > > > > > > > > > > > > > > > > > > > > > Would be very grateful if we could get to 3.0 since it > > is the new feature > > > > > > > > > > > of the delayed message and has no impact on other > > components. > > > > > > > > > > > > > > > > > > > > > > These PR is important to PIP-195, they will fix some > > problem with the new > > > > > > > > > > > delayed message and make this new feature work better. > > > > > > > > > > > > > > > > > > > > > > Need to cherry-pick PR: > > > > > > > > > > > https://github.com/apache/pulsar/pull/20086 > > > > > > > > > > > https://github.com/apache/pulsar/pull/20111 > > > > > > > > > > > https://github.com/apache/pulsar/pull/20126 > > > > > > > > > > > https://github.com/apache/pulsar/pull/20136 > > > > > > > > > > > https://github.com/apache/pulsar/pull/20117 > > > > > > > > > > > https://github.com/apache/pulsar/pull/20155 > > > > > > > > > > > https://github.com/apache/pulsar/pull/20156 > > > > > > > > > > > https://github.com/apache/pulsar/pull/20158 > > > > > > > > > > > > > > > > > > > > > > Thanks, > > > > > > > > > > > Cong Zhao > > > > > > > > > > > > > > > > > > > > > > On 2023/04/21 05:36:26 Zike Yang wrote: > > > > > > > > > > > > Hi, all > > > > > > > > > > > > > > > > > > > > > > > > We found that there were a lot of commits that were > > cherry-picked to > > > > > > > > > > > > branch-3.0 without any discussion or notification to > > reach a > > > > > > > > > > > > consensus. Some PRs marked for the 3.1.0 milestone > > were also > > > > > > > > > > > > cherry-picked into branch-3.0. Here are the > > corresponding > > > > > > > > > > > > cherry-picked commits: > > > > > > > > > > > > * > > > > > > > > > > > > > > https://github.com/apache/pulsar/commit/3da39b2e1553cecbc6d6b85e8bc7844f611d5637 > > > > > > > > > > > > * > > > > > > > > > > > > > > https://github.com/apache/pulsar/commit/e248e14473142765db1df324c20a081a2422980e > > > > > > > > > > > > * > > > > > > > > > > > > > > https://github.com/apache/pulsar/commit/e27abe9e128fb71b65ffe06417574c9a7f3facbd > > > > > > > > > > > > * > > > > > > > > > > > > > > https://github.com/apache/pulsar/commit/e1d63990644700bf61b3d7af1ef6d4d62145c2bb > > > > > > > > > > > > * > > > > > > > > > > > > > > https://github.com/apache/pulsar/commit/ff59240165c73a9c3a3dcca20702ab44b0b18d33 > > > > > > > > > > > > * > > > > > > > > > > > > > > https://github.com/apache/pulsar/commit/49480ea558e647169e8df01bfd2e871a5386e19e > > > > > > > > > > > > * > > > > > > > > > > > > > > https://github.com/apache/pulsar/commit/1d1a3ef864a65c995ceda4b7875ed934c2574298 > > > > > > > > > > > > * > > > > > > > > > > > > > > https://github.com/apache/pulsar/commit/0df741259505b9189147d72c6f013b2c18f0436c > > > > > > > > > > > > * > > > > > > > > > > > > > > https://github.com/apache/pulsar/commit/d05871213adc351d4c718c2a6fb0909b01d279ff > > > > > > > > > > > > > > > > > > > > > > > > According to our release policy[0] and the code > freeze > > notification > > > > > > > > > > > > [1], all commits that need to be cherry-picked into > > branch-3.0 will > > > > > > > > > > > > require reaching the consensus before cherry-picking. > > > > > > > > > > > > Before cherry-picking the commit, we need to provide > > the context for > > > > > > > > > > > > why it needs to be cherry-picked into branch-3.0. > Then > > mark it with > > > > > > > > > > > > the 3.0.0 milestone and raise the discussion to reach > > a consensus. > > > > > > > > > > > > > > > > > > > > > > > > I would like to start a discussion regarding the > > context for the above > > > > > > > > > > > > cherry-picked commits. Should we include them in > > Pulsar 3.0? > > > > > > > > > > > > If there is still no consensus for the above commits, > > we may need to > > > > > > > > > > > > revert them before the RC3. > > > > > > > > > > > > > > > > > > > > > > > > If you have any questions or concerns, please do not > > hesitate to let > > > > > > > > > > > > us know. Thanks for your cooperation. > > > > > > > > > > > > > > > > > > > > > > > > [0] > > https://pulsar.apache.org/contribute/release-policy/#release-cycles > > > > > > > > > > > > [1] > > https://lists.apache.org/thread/43d28rtzsx7x3o4zd523jr5dmnczrn4h > > > > > > > > > > > > > > > > > > > > > > > > Thanks, > > > > > > > > > > > > Zike Yang > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > >
