You are viewing a plain text version of this content. The canonical link for it is here.
Posted to dev@cassandra.apache.org by Bernardo Botella <co...@bernardobotella.com> on 2026/02/13 15:43:30 UTC
[DISCUSS] Backport async profiling to all GA branches - CASSSIDECAR-254
Hi everyone!
We added async profiling to Cassandra via nodetool command on this commit:
https://github.com/apache/cassandra/commit/7c3c3a1d86782a515583f89c6f17fb30e7f5e41e
With a bug fix in this commit:
https://github.com/apache/cassandra/commit/d1b4ddc32acd8328f83ea7b5dd5a8cd3437ded04
Fix JMXFeatureTest failure due to disabled AsyncProfiler · apache/cassandra@d1b4ddc
github.com
I think this feature will be useful for other branches as well. It is pretty self contained, and I see low risk on back porting it.
I wanted to pick the community collective brain on this one to understand if there would be any concerns for this to be back ported, or I can start working on it.
Thanks!
Bernardo
Re: [DISCUSS] Backport async profiling to all GA branches - CASSSIDECAR-254
Posted by Brad <bs...@gmail.com>.
I also agree that it would be very very helpful to backport this to at
least 5.0.x.
On Fri, Feb 13, 2026 at 5:23 PM Isaac Reath <is...@gmail.com> wrote:
> I'd be in favor of backporting this. Looking at the patch, it looks well
> isolated and guarded behind feature flags and I agree that it's very useful
> for other branches.
>
> On Fri, Feb 13, 2026 at 10:44 AM Bernardo Botella <
> contacto@bernardobotella.com> wrote:
>
>> Hi everyone!
>>
>> We added async profiling to Cassandra via nodetool command on this commit:
>> Initial async-profiler Nodetool implementation · apache/cassandra@7c3c3a1
>> <https://github.com/apache/cassandra/commit/7c3c3a1d86782a515583f89c6f17fb30e7f5e41e>
>> github.com
>> <https://github.com/apache/cassandra/commit/7c3c3a1d86782a515583f89c6f17fb30e7f5e41e>
>> [image: apple-touch-icon-180x180-a80b8e11abe2.png]
>> <https://github.com/apache/cassandra/commit/7c3c3a1d86782a515583f89c6f17fb30e7f5e41e>
>> <https://github.com/apache/cassandra/commit/7c3c3a1d86782a515583f89c6f17fb30e7f5e41e>
>>
>>
>> With a bug fix in this commit:
>> [image: d1b4ddc32acd8328f83ea7b5dd5a8cd3437ded04.png]
>>
>> Fix JMXFeatureTest failure due to disabled AsyncProfiler ·
>> apache/cassandra@d1b4ddc
>> <https://github.com/apache/cassandra/commit/d1b4ddc32acd8328f83ea7b5dd5a8cd3437ded04>
>> github.com
>> <https://github.com/apache/cassandra/commit/d1b4ddc32acd8328f83ea7b5dd5a8cd3437ded04>
>>
>> <https://github.com/apache/cassandra/commit/d1b4ddc32acd8328f83ea7b5dd5a8cd3437ded04>
>>
>> I think this feature will be useful for other branches as well. It is
>> pretty self contained, and I see low risk on back porting it.
>>
>> I wanted to pick the community collective brain on this one to understand
>> if there would be any concerns for this to be back ported, or I can start
>> working on it.
>>
>> Thanks!
>> Bernardo
>>
>
Re: [DISCUSS] Backport async profiling to all GA branches - CASSSIDECAR-254
Posted by Isaac Reath <is...@gmail.com>.
I'd be in favor of backporting this. Looking at the patch, it looks well
isolated and guarded behind feature flags and I agree that it's very useful
for other branches.
On Fri, Feb 13, 2026 at 10:44 AM Bernardo Botella <
contacto@bernardobotella.com> wrote:
> Hi everyone!
>
> We added async profiling to Cassandra via nodetool command on this commit:
> Initial async-profiler Nodetool implementation · apache/cassandra@7c3c3a1
> <https://github.com/apache/cassandra/commit/7c3c3a1d86782a515583f89c6f17fb30e7f5e41e>
> github.com
> <https://github.com/apache/cassandra/commit/7c3c3a1d86782a515583f89c6f17fb30e7f5e41e>
> [image: apple-touch-icon-180x180-a80b8e11abe2.png]
> <https://github.com/apache/cassandra/commit/7c3c3a1d86782a515583f89c6f17fb30e7f5e41e>
> <https://github.com/apache/cassandra/commit/7c3c3a1d86782a515583f89c6f17fb30e7f5e41e>
>
>
> With a bug fix in this commit:
> [image: d1b4ddc32acd8328f83ea7b5dd5a8cd3437ded04.png]
>
> Fix JMXFeatureTest failure due to disabled AsyncProfiler ·
> apache/cassandra@d1b4ddc
> <https://github.com/apache/cassandra/commit/d1b4ddc32acd8328f83ea7b5dd5a8cd3437ded04>
> github.com
> <https://github.com/apache/cassandra/commit/d1b4ddc32acd8328f83ea7b5dd5a8cd3437ded04>
>
> <https://github.com/apache/cassandra/commit/d1b4ddc32acd8328f83ea7b5dd5a8cd3437ded04>
>
> I think this feature will be useful for other branches as well. It is
> pretty self contained, and I see low risk on back porting it.
>
> I wanted to pick the community collective brain on this one to understand
> if there would be any concerns for this to be back ported, or I can start
> working on it.
>
> Thanks!
> Bernardo
>
Re: [DISCUSS] Backport async profiling to all GA branches - CASSSIDECAR-254
Posted by Isaac Reath <is...@gmail.com>.
I'd be happy to volunteer to take a backport of this feature out to
production if that helps push this more towards being "prod-ready." I see
enough value in this feature that I would want to either way, even if we
ultimately decide to not merge it into a stable branch. That said, I can
still be sympathetic to how this is different from auto-repair in terms of
demand for the feature if that ultimately is the deciding factor for
whether or not to backport.
On Fri, Feb 20, 2026 at 10:04 AM Josh McKenzie <jm...@apache.org> wrote:
> Fair point on it not being widely used in a prod environment yet (unless
> someone wants to step forward and advocate for it). It does open the door
> for questions around our stance on backporting isolated features flagged as
> experimental to existing GA branches and whether that's any different than
> backporting features we advertise as prod ready.
>
> I really don't want us to regress when it comes to our stance on stability
> as being priority #1, but our releases are very long-lived and some of
> these things we're adding to GA branches are kind of "ecosystem technical
> debt", in that we're some years behind on making industry standard features
> and tooling available to users.
>
> On Thu, Feb 19, 2026, at 6:33 PM, Štefan Miklošovič wrote:
>
> Up to you guys! I believe your judgement here. I am just saying, as a
> fact, that I am not sure there are a lot of precedents when it comes to
> backporting a feature which was not released yet.
>
> Repairs are something else. It runs in production already, I believe, for
> some users. I am just not 100% comfortable with backporting something we do
> not know how it will be received or if it has some gaps or issues. We think
> it does not.
>
> Anyway proceed as you see fit :) I am only happy that you think we did
> such a great feature that it begs for backporting even if it was not
> released yet ..
>
> On Thu, Feb 19, 2026 at 2:48 PM Josh McKenzie <jm...@apache.org>
> wrote:
>
>
> Flipping through the PR, it looks like the vast majority of the 2k patch
> is test code and docs with some very small (10 lines or so) additions of
> API endpoints to various services + the new async service at ~ 500. This
> looks like a backport candidate with very low risk to users who don't
> enable it.
>
> We could backport it and doc / flag it as an experimental feature on older
> branches if we're worried the profiler integration when combined w/the C*
> code is unstable or could cause slowdown or outage.
>
> I'm +1 on backporting, marked experimental or not. Seems useful and very
> low risk.
>
> On Wed, Feb 18, 2026, at 11:15 PM, Štefan Miklošovič wrote:
>
> I would wait until 6.0 is out and it has more exposure. It has to be used
> for a while so we are confident and comfortable that backporting is okay.
> Maybe some issues or improvements will be done and then we would need to
> backport that too? Let's just mature it first before making that jump.
>
> On Fri, Feb 13, 2026 at 4:44 PM Bernardo Botella <
> contacto@bernardobotella.com> wrote:
>
> Hi everyone!
>
> We added async profiling to Cassandra via nodetool command on this commit:
> Initial async-profiler Nodetool implementation · apache/cassandra@7c3c3a1
> <https://github.com/apache/cassandra/commit/7c3c3a1d86782a515583f89c6f17fb30e7f5e41e>
> github.com
> <https://github.com/apache/cassandra/commit/7c3c3a1d86782a515583f89c6f17fb30e7f5e41e>
> [image: apple-touch-icon-180x180-a80b8e11abe2.png]
> <https://github.com/apache/cassandra/commit/7c3c3a1d86782a515583f89c6f17fb30e7f5e41e>
>
> With a bug fix in this commit:
> [image: d1b4ddc32acd8328f83ea7b5dd5a8cd3437ded04.png]
> Fix JMXFeatureTest failure due to disabled AsyncProfiler ·
> apache/cassandra@d1b4ddc
> <https://github.com/apache/cassandra/commit/d1b4ddc32acd8328f83ea7b5dd5a8cd3437ded04>
> github.com
> <https://github.com/apache/cassandra/commit/d1b4ddc32acd8328f83ea7b5dd5a8cd3437ded04>
> I think this feature will be useful for other branches as well. It is
> pretty self contained, and I see low risk on back porting it.
>
> I wanted to pick the community collective brain on this one to understand
> if there would be any concerns for this to be back ported, or I can start
> working on it.
>
> Thanks!
> Bernardo
>
>
>
>
Re: [DISCUSS] Backport async profiling to all GA branches - CASSSIDECAR-254
Posted by Josh McKenzie <jm...@apache.org>.
Fair point on it not being widely used in a prod environment yet (unless someone wants to step forward and advocate for it). It does open the door for questions around our stance on backporting isolated features flagged as experimental to existing GA branches and whether that's any different than backporting features we advertise as prod ready.
I really don't want us to regress when it comes to our stance on stability as being priority #1, but our releases are very long-lived and some of these things we're adding to GA branches are kind of "ecosystem technical debt", in that we're some years behind on making industry standard features and tooling available to users.
On Thu, Feb 19, 2026, at 6:33 PM, Štefan Miklošovič wrote:
> Up to you guys! I believe your judgement here. I am just saying, as a fact, that I am not sure there are a lot of precedents when it comes to backporting a feature which was not released yet.
>
> Repairs are something else. It runs in production already, I believe, for some users. I am just not 100% comfortable with backporting something we do not know how it will be received or if it has some gaps or issues. We think it does not.
>
> Anyway proceed as you see fit :) I am only happy that you think we did such a great feature that it begs for backporting even if it was not released yet ..
>
> On Thu, Feb 19, 2026 at 2:48 PM Josh McKenzie <jm...@apache.org> wrote:
>> __
>> Flipping through the PR, it looks like the vast majority of the 2k patch is test code and docs with some very small (10 lines or so) additions of API endpoints to various services + the new async service at ~ 500. This looks like a backport candidate with very low risk to users who don't enable it.
>>
>> We could backport it and doc / flag it as an experimental feature on older branches if we're worried the profiler integration when combined w/the C* code is unstable or could cause slowdown or outage.
>>
>> I'm +1 on backporting, marked experimental or not. Seems useful and very low risk.
>>
>> On Wed, Feb 18, 2026, at 11:15 PM, Štefan Miklošovič wrote:
>>> I would wait until 6.0 is out and it has more exposure. It has to be used for a while so we are confident and comfortable that backporting is okay. Maybe some issues or improvements will be done and then we would need to backport that too? Let's just mature it first before making that jump.
>>>
>>> On Fri, Feb 13, 2026 at 4:44 PM Bernardo Botella <co...@bernardobotella.com> wrote:
>>>> Hi everyone!
>>>>
>>>> We added async profiling to Cassandra via nodetool command on this commit:
>>>> Initial async-profiler Nodetool implementation · apache/cassandra@7c3c3a1 <https://github.com/apache/cassandra/commit/7c3c3a1d86782a515583f89c6f17fb30e7f5e41e>
>>>> github.com <https://github.com/apache/cassandra/commit/7c3c3a1d86782a515583f89c6f17fb30e7f5e41e>
>>>> apple-touch-icon-180x180-a80b8e11abe2.png <https://github.com/apache/cassandra/commit/7c3c3a1d86782a515583f89c6f17fb30e7f5e41e>
>>>>
>>>> With a bug fix in this commit:
>>>> d1b4ddc32acd8328f83ea7b5dd5a8cd3437ded04.png
>>>> Fix JMXFeatureTest failure due to disabled AsyncProfiler · apache/cassandra@d1b4ddc <https://github.com/apache/cassandra/commit/d1b4ddc32acd8328f83ea7b5dd5a8cd3437ded04>
>>>> github.com <https://github.com/apache/cassandra/commit/d1b4ddc32acd8328f83ea7b5dd5a8cd3437ded04>
>>>> I think this feature will be useful for other branches as well. It is pretty self contained, and I see low risk on back porting it.
>>>>
>>>> I wanted to pick the community collective brain on this one to understand if there would be any concerns for this to be back ported, or I can start working on it.
>>>>
>>>> Thanks!
>>>> Bernardo
>>>>
>>
Re: [DISCUSS] Backport async profiling to all GA branches - CASSSIDECAR-254
Posted by Štefan Miklošovič <sm...@apache.org>.
Up to you guys! I believe your judgement here. I am just saying, as a fact,
that I am not sure there are a lot of precedents when it comes to
backporting a feature which was not released yet.
Repairs are something else. It runs in production already, I believe, for
some users. I am just not 100% comfortable with backporting something we do
not know how it will be received or if it has some gaps or issues. We think
it does not.
Anyway proceed as you see fit :) I am only happy that you think we did such
a great feature that it begs for backporting even if it was not released
yet ..
On Thu, Feb 19, 2026 at 2:48 PM Josh McKenzie <jm...@apache.org> wrote:
> Flipping through the PR, it looks like the vast majority of the 2k patch
> is test code and docs with some very small (10 lines or so) additions of
> API endpoints to various services + the new async service at ~ 500. This
> looks like a backport candidate with very low risk to users who don't
> enable it.
>
> We could backport it and doc / flag it as an experimental feature on older
> branches if we're worried the profiler integration when combined w/the C*
> code is unstable or could cause slowdown or outage.
>
> I'm +1 on backporting, marked experimental or not. Seems useful and very
> low risk.
>
> On Wed, Feb 18, 2026, at 11:15 PM, Štefan Miklošovič wrote:
>
> I would wait until 6.0 is out and it has more exposure. It has to be used
> for a while so we are confident and comfortable that backporting is okay.
> Maybe some issues or improvements will be done and then we would need to
> backport that too? Let's just mature it first before making that jump.
>
> On Fri, Feb 13, 2026 at 4:44 PM Bernardo Botella <
> contacto@bernardobotella.com> wrote:
>
> Hi everyone!
>
> We added async profiling to Cassandra via nodetool command on this commit:
> Initial async-profiler Nodetool implementation · apache/cassandra@7c3c3a1
> <https://github.com/apache/cassandra/commit/7c3c3a1d86782a515583f89c6f17fb30e7f5e41e>
> github.com
> <https://github.com/apache/cassandra/commit/7c3c3a1d86782a515583f89c6f17fb30e7f5e41e>
> [image: apple-touch-icon-180x180-a80b8e11abe2.png]
> <https://github.com/apache/cassandra/commit/7c3c3a1d86782a515583f89c6f17fb30e7f5e41e>
>
> With a bug fix in this commit:
> [image: d1b4ddc32acd8328f83ea7b5dd5a8cd3437ded04.png]
> Fix JMXFeatureTest failure due to disabled AsyncProfiler ·
> apache/cassandra@d1b4ddc
> <https://github.com/apache/cassandra/commit/d1b4ddc32acd8328f83ea7b5dd5a8cd3437ded04>
> github.com
> <https://github.com/apache/cassandra/commit/d1b4ddc32acd8328f83ea7b5dd5a8cd3437ded04>
> I think this feature will be useful for other branches as well. It is
> pretty self contained, and I see low risk on back porting it.
>
> I wanted to pick the community collective brain on this one to understand
> if there would be any concerns for this to be back ported, or I can start
> working on it.
>
> Thanks!
> Bernardo
>
>
>
Re: [DISCUSS] Backport async profiling to all GA branches - CASSSIDECAR-254
Posted by Josh McKenzie <jm...@apache.org>.
Flipping through the PR, it looks like the vast majority of the 2k patch is test code and docs with some very small (10 lines or so) additions of API endpoints to various services + the new async service at ~ 500. This looks like a backport candidate with very low risk to users who don't enable it.
We could backport it and doc / flag it as an experimental feature on older branches if we're worried the profiler integration when combined w/the C* code is unstable or could cause slowdown or outage.
I'm +1 on backporting, marked experimental or not. Seems useful and very low risk.
On Wed, Feb 18, 2026, at 11:15 PM, Štefan Miklošovič wrote:
> I would wait until 6.0 is out and it has more exposure. It has to be used for a while so we are confident and comfortable that backporting is okay. Maybe some issues or improvements will be done and then we would need to backport that too? Let's just mature it first before making that jump.
>
> On Fri, Feb 13, 2026 at 4:44 PM Bernardo Botella <co...@bernardobotella.com> wrote:
>> Hi everyone!
>>
>> We added async profiling to Cassandra via nodetool command on this commit:
>> Initial async-profiler Nodetool implementation · apache/cassandra@7c3c3a1 <https://github.com/apache/cassandra/commit/7c3c3a1d86782a515583f89c6f17fb30e7f5e41e>
>> github.com <https://github.com/apache/cassandra/commit/7c3c3a1d86782a515583f89c6f17fb30e7f5e41e>
>> apple-touch-icon-180x180-a80b8e11abe2.png <https://github.com/apache/cassandra/commit/7c3c3a1d86782a515583f89c6f17fb30e7f5e41e>
>>
>> With a bug fix in this commit:
>> d1b4ddc32acd8328f83ea7b5dd5a8cd3437ded04.png
>> Fix JMXFeatureTest failure due to disabled AsyncProfiler · apache/cassandra@d1b4ddc <https://github.com/apache/cassandra/commit/d1b4ddc32acd8328f83ea7b5dd5a8cd3437ded04>
>> github.com <https://github.com/apache/cassandra/commit/d1b4ddc32acd8328f83ea7b5dd5a8cd3437ded04>
>> I think this feature will be useful for other branches as well. It is pretty self contained, and I see low risk on back porting it.
>>
>> I wanted to pick the community collective brain on this one to understand if there would be any concerns for this to be back ported, or I can start working on it.
>>
>> Thanks!
>> Bernardo
>>
Re: [DISCUSS] Backport async profiling to all GA branches - CASSSIDECAR-254
Posted by Štefan Miklošovič <sm...@apache.org>.
I would wait until 6.0 is out and it has more exposure. It has to be used
for a while so we are confident and comfortable that backporting is okay.
Maybe some issues or improvements will be done and then we would need to
backport that too? Let's just mature it first before making that jump.
On Fri, Feb 13, 2026 at 4:44 PM Bernardo Botella <
contacto@bernardobotella.com> wrote:
> Hi everyone!
>
> We added async profiling to Cassandra via nodetool command on this commit:
> Initial async-profiler Nodetool implementation · apache/cassandra@7c3c3a1
> <https://github.com/apache/cassandra/commit/7c3c3a1d86782a515583f89c6f17fb30e7f5e41e>
> github.com
> <https://github.com/apache/cassandra/commit/7c3c3a1d86782a515583f89c6f17fb30e7f5e41e>
> [image: apple-touch-icon-180x180-a80b8e11abe2.png]
> <https://github.com/apache/cassandra/commit/7c3c3a1d86782a515583f89c6f17fb30e7f5e41e>
> <https://github.com/apache/cassandra/commit/7c3c3a1d86782a515583f89c6f17fb30e7f5e41e>
>
>
> With a bug fix in this commit:
> [image: d1b4ddc32acd8328f83ea7b5dd5a8cd3437ded04.png]
>
> Fix JMXFeatureTest failure due to disabled AsyncProfiler ·
> apache/cassandra@d1b4ddc
> <https://github.com/apache/cassandra/commit/d1b4ddc32acd8328f83ea7b5dd5a8cd3437ded04>
> github.com
> <https://github.com/apache/cassandra/commit/d1b4ddc32acd8328f83ea7b5dd5a8cd3437ded04>
>
> <https://github.com/apache/cassandra/commit/d1b4ddc32acd8328f83ea7b5dd5a8cd3437ded04>
>
> I think this feature will be useful for other branches as well. It is
> pretty self contained, and I see low risk on back porting it.
>
> I wanted to pick the community collective brain on this one to understand
> if there would be any concerns for this to be back ported, or I can start
> working on it.
>
> Thanks!
> Bernardo
>