Hi Nick,
Thanks for the KIP.
Sorry for the late discussion. I was reading the KIP, nudged by the voting
thread and I
have a couple of comments.
AS1: The old syntax should eventually not be supported. Please could the KIP say
something like "The previous syntax using positional arguments is deprecated
and will
be removed from the parser in the next major Kafka release. A deprecation
warning message
will be displayed if positional arguments are used to encourage users to move
to the
new syntax." And then you should remove the old syntax in AK 5.0.
AS2: The producer perf test uses slightly different names for the same
parameters. I
suggest you align with it, and the conventions of other similar tools.
* --num-records instead of --num-messages
* --record-size instead of --message-size-bytes
* --command-config instance of --properties-file (used by many tools)
* How about --record-key-size instead of --message-key-size-bytes
* How about --record-header-size instead of --message-header-size-bytes
Thanks,
Andrew
________________________________________
From: Junwang Guo <lansg0...@gmail.com>
Sent: 08 July 2025 09:33
To: dev@kafka.apache.org <dev@kafka.apache.org>
Subject: Re: [DISCUSS] KIP-1172: Improve EndToEndLatency tool
Hi everyone,
Just giving this thread a gentle nudge. If there's no further feedback, I
plan to start the vote tomorrow.
Regards,
Nick
On Wed, Jun 11, 2025 at 6:04 PM Junwang Guo <lansg0...@gmail.com> wrote:
> Hi Chia-Ping
>
> Thanks for the feedback!
>
>
> chia_0: Good suggestion! I've expanded the "Compatibility" section to
> include more detailed e2e testing descriptions.
>
> chia_1: Exactly—the new fields only work with the new argument parser.
> I've made this explicit in the KIP to avoid any confusion.
>
> chia_2: If “key” and “header” are added, we will apply the same check as
> we do for the record “value,” for example:
>
>
> String sentKey = new String(sentMessageKey, StandardCharsets.UTF_8);
>
> String readKey = new String(records.iterator().next().key(),
> StandardCharsets.UTF_8);
>
> Header sentHeader = headers.iterator().next();
> Header readHeader = records.iterator().next().headers().iterator().next();
>
> if (!readHeader.equals(sentHeader)) {
> throw new RuntimeException();
>
> if (!readKey.equals(sentKey))
> throw new RuntimeException();
> }
>
>
> Regards,
>
> Nick
>
> On Tue, Jun 10, 2025 at 10:55 PM Chia-Ping Tsai <chia7...@gmail.com>
> wrote:
>
>> hi Nick
>>
>> thanks for this proposal. Some questions are listed below.
>>
>> chia_0: This tool is used by e2e, so could you please describe the changes
>> for e2e too?
>>
>> chia_1: the two new fields (*message-key-size-byt and *
>> *message-header-size-bytes)* are NOT supported by old (index) arguments,
>> right? If so, do you mind mentioning that in the KIP?
>>
>> chia_2: `EndToEndLatency` will validate the record value. What happens if
>> the "key" and "header" are added to the record?
>>
>> Best,
>> Chia-Ping
>>
>> Junwang Guo <lansg0...@gmail.com> 於 2025年6月10日 週二 下午10:29寫道：
>>
>> > Hi everyone,
>> >
>> > I would like to start a discussion on a KIP to improve the
>> > `EndToEndLatency` tool.
>> >
>> > KIP Link: https://cwiki.apache.org/confluence/x/Awu9F
>> >
>> > Thank you!
>> >
>> > Best regards,
>> > Nick Guo
>> >
>>
>
