Hi,
IMO, there are two scenarios when users need test coverage:
1. When new features or major optimizations are submitted, the owners of
the PR may care about the code coverage of these new features.
2. When we need to optimize the original code to increase the overall code
test coverage of Pulsar.
The pulsar-client-reactive repository has been created.
I have added the first few commits as a starting point for the library.
The snapshot version of the library has been published to the Apache snapshots
repository.
Please fork and star this repository:
https://github.com/apache/pulsar-client-
Hi,
IMO, code coverage is not just a number, 50% or 70% makes no sense except
to let us feel that we have more confidence. So what is really important? I
think it is the *coverage* itself, we need to see where we need to write
tests in the future based the result because only if we have this data,
> Please provide data point its impact to CI stability.
If you take a look at https://github.com/apache/pulsar/pull/17382, it
changes pulsar-ci.yml to run commands for generating the codecov report.
I'm unsure of the impact of a new step, it may not affect too much since
there's already a runner t
Hi Tison,
Thanks for your input. I agree that the community should focus on the
priorities regarding CI that you mentioned. At the same time, I'm having a
hard time understanding the negative impact that you suggested from this
change.
1. To my knowledge code coverage calculation adds little over
Hi Lin,
Thanks for starting this discussion!
As long as it takes a different resource set from current CI tasks, I'm +0
as commented on PR-17382. I hardly read the report.
I read the output in your proposal as simply:
> The report will serve as additional input for the reviewers. The requester
Hi,
I'd like to start a discussion about turning on CodeCov report for PRs to
master to show the PR's impact on unit test coverage. Previous discussion
on https://github.com/apache/pulsar/pull/17382.
Proposal:
1. Unit test coverage will be added to the CI pipeline and reported to the
PR page.
Sam
Hi Mingyu
> In the exclusive consumption mode of exclusive, failover, reader, etc.
If I understand correctly, your concern is not about read-only partitions,
but about being able to read newly sent messages as quickly as possible.
Maybe this problem can be solved by adding a new configuration that
Hi, Bo
I totally agree with this approach.
Suppose we now implement deduplication for common messages on the client
side. In that case, there is no need to add other logic that may cause API
break-changes to guarantee transaction exactly-once semantics.
Yours,
Xiangying
On Fri, Sep 9, 2022 at 10:1
+1
This suggestion is really awesome.
The new solution can reduce the learning cost of Pulsar users and avoid
them taking many detours.
Yours,
Xiangying
On Fri, Sep 9, 2022 at 8:05 PM Shivji Kumar Jha wrote:
> Created https://github.com/apache/pulsar/issues/17574 for Yunze's
> suggestion on expo
Hi all,
you can see the problem in a google doc and comments.
google doc link:
https://docs.google.com/document/d/1J1xGcj8YORrdlCa_XDt28uV0TMp03gSmX_z43ZRhwZo/edit?usp=sharing
Thanks!
Bo
丛搏 于2022年9月8日周四 10:55写道:
>
> Hello, Pulsar community:
>
>
> Now the consumer does not filter messages that
Hi Michael
you are right, now the current handover is very blunt. We can
implement a flexible and efficient solution by changing the protocol
in a 3.0 enhancement. But we also need this optimization in 2.0, but
this optimization should not be too complicated at best.
Thanks,
Bo
Michael Marshall
Hi Haiting,
good point! we need to pay attention to all reset cursor operations. I
think this is a configurable optimization because some users don't
call `void redeliverUnacknowledgedMessages().` This optimization is a
break-change operation.
I will find all the operations to reset the cursor, an
Created https://github.com/apache/pulsar/issues/17574 for Yunze's
suggestion on exposing the following in java client
void acknowledgeCumulative(Map topicToMessageId);
Regards,
Shivji Kumar Jha
http://www.shivjijha.com/
+91 8884075512
On Fri, 9 Sept 2022 at 11:44, Shivji Kumar Jha wrote:
> T
tisonkun commented on PR #70:
URL: https://github.com/apache/pulsar-test-infra/pull/70#issuecomment-1241891172
@nodece since the manner changed, I think we should update the comment
content also for simply inlining the information to ask the user to add a valid
checkbox, instead of linking
+1(binding)
Great work !
Regards
Jiwei Guo (Tboy)
On Fri, Sep 9, 2022 at 10:10 AM PengHui Li wrote:
> +1(binding)
>
> I have done the review on gdoc
> And please also update the github issue(PIP).
>
> Thanks,
> Penghui
>
> On Fri, Sep 9, 2022 at 9:31 AM 丛搏 wrote:
>
> > Hi, Xiangying
> > +1(
As you may have noticed, the CI is slow again.
There are more than 140 workflows pending:
https://github.com/apache/pulsar/actions?query=is%3Aqueued
There are only 2-3 workflows in progress:
https://github.com/apache/pulsar/actions?query=is%3Ain_progress
Lari and I believe that we're still penaliz
Yes ,it is a new message router using partition permission without the
partition key .
On 2022/08/29 01:43:20 PengHui Li wrote:
> Does it look like a new message router for the case that the message is
> without the partition key?
> The only difference is the router is based on the backlog.
>
>
Done
Nicolò Boschi
Il giorno ven 9 set 2022 alle ore 08:32 Nicolò Boschi
ha scritto:
> That's a good observation. We don't have data so we can't be sure about
> that, but I think that it can be a possible cause of the recent congestion.
>
> I opened an INFRA ticket -
> https://issues.apache.
19 matches
Mail list logo