Penghui,

I didn't know that there were so many known bugs around transactions
scheduled for 2.9.3, my bad.

However, as Enrico pointed out, the issue impacts Pulsar clients that are
not using the transactions, so we can't just say - ok, just another bug
about transactions, it's not critical since they're not production ready
(btw, where we state that they aren't production ready on the
documentation?).


The workaround you mentioned is not always viable, since you can have
clients of different tenants/customers that are not using transactions
while, at the same time, a little portion that are experiencing with them.

I agree that it is uncommon to have only one message produced. On the other
hand, it's a very common case where other projects using Pulsar have
unit/integration tests that write only one message and expect to be
consumed (that's because they test the application logic and not Pulsar).


Given that, it's fair to say that 2.9.2 is not worse than 2.9.1, so,
finally, we can go ahead.

Looking forward to see 2.9.3 soon


I tested the artifacts, so I'll put my vote here:


+1 (non binding)


Checks:

- Checksum and signatures

- Apache Rat check passes

- Compile from source w JDK11

- Build docker image from source

- Run Pulsar standalone and produce-consume from CLI


BR,

Nicolò

Il giorno gio 10 feb 2022 alle ore 13:39 PengHui Li <peng...@apache.org> ha
scritto:

> > Please go ahead with the release, I won't VOTE on this thread.
> But I hope we can follow up soon with a new release, otherwise due to that
> bug
> you cannot enable transactions on your Pulsar cluster if you have to
> support Pulsar client that do not enable transactions
>
>
> Yes, agree. We will follow up the 2.9.3 soon. There are other
> ongoing transaction fixes
> we will complete them ASAP and provide a version with certain guarantees
> for transaction stability.
> We are doing lots of tests these days, 2.9.3 should be a good version for
> transactions.
>
> Thanks,
> Penghui
>
>
> On Thu, Feb 10, 2022 at 7:37 PM Lin Lin <lin...@apache.org> wrote:
>
> >
> >
> > +1(binding)
> >
> > 1. Checked the signature
> > 2. Start standalone
> > 3. Publish and consume successfully
> > 4. Checked function
> >
>


-- 
Nicolò Boschi

Reply via email to