> And to nobody's surprise, we are suffering with PRs. 

Glad to see the data goes back to 2011… what I see is we float around .6 to .8 
for "Change Request Closure Ratio (GitHub PR)”… and AI didn’t change anything 
here?  At least looking at the last 2 years as a proxy, this metric doesn’t 
really look any different from between 2020 to now.

"Open PR backlog (total)” also reenforces this.  The other projects (Kafka, 
Spark, Flink, and DataFusion) all have a clear hockey stick pattern starting 
2024 but Cassandra doesn’t; the pattern is linear growth…

What is more telling is "Contributor Absence Factor”, Cassandra is at 7 which 
is basically unchanged since 2015.  And from 2022 Kafka has gone from ~8 to 
25...

> On Oct 13, 2026, at 11:07 AM, Patrick McFadin <[email protected]> wrote:
> 
> Thanks for the feedback I got so far. As a result of one of those discussions 
> I added every sub-project to the project health page. You can now select 
> which project to look at and it will give you a similar page of stats for 
> Cassandra or each sub-project.  
> 
> Patrick
> 
>> On Oct 10, 2026, at 11:10 PM, Patrick McFadin <[email protected]> wrote:
>> 
>> Hello from sunny and warm Glasgow and Community Over Code. 
>> 
>> A series of things lately, have me asking questions about how our community 
>> is doing, how the project is doing and is this something special to the 
>> Cassandra project? Not an easy thing to measure so I decided to find out if 
>> it was possible. 
>> 
>> The answer is yes! Linux Foundation already does a health status page for 
>> their projects, with lots of metrics and with an entire subproject called 
>> CHAOSS[1] This lead me to develop this first version of our project health 
>> geared towards apache projects:
>> 
>> https://pmcfadin.github.io/cassandra-project-health/index.html
>> 
>> There are a lot of metrics, but they are grouped into, 
>>  Community: Code and contributor health metrics
>>  Governance:  How each commit stands up to our standards of testing and 
>> review
>>  Conversations: Sentiment analysis across every email in dev@ and every JIRA 
>> comment
>>  Peers: Select group of metrics form other projects as a comparison
>> 
>> A couple things stand out very quickly. We are very responsive on mailing 
>> lists and great with new people. And to nobody's surprise, we are suffering 
>> with PRs. 
>> 
>> That lead me to another path. In teams I work with, automated review 
>> pipelines are getting really deep with AI assitance and it's needed. There 
>> is a lot of toil just opening a PR to look at and start a review process. 
>> I've polled collegues, used existing skills like the rustrazorblade 
>> skills[2] and practices I've seen from my own work. 
>> 
>> https://pmcfadin.github.io/cassandra-pr-review/
>> 
>> This is a limited group of PRs I did a a POC but ideally, this would be run 
>> as soon as a new PR is recived and the contributor would get a report link 
>> within minutes. I also did not every single script in my pipeline for every 
>> PR. A good example of every script being run is PR 5021: 
>> 
>> https://pmcfadin.github.io/cassandra-pr-review/pr/5201/
>> 
>> Some highlights of what it does.
>>  Check to make sure all elements are present. Tests. Reviewer. CI results. 
>>  Looking at the code changes and suggesting reviewers
>>  Adversarial code review output
>>  Static analysis using PMD
>>  A test plan for if you want to do deeper testing using Easy DB Lab
>>  A lot lot more...
>> 
>> For new contributors, this gives them immediate feedback and gives them a 
>> self-service aspect. You can run all these locally as well. 
>> For reviewers, you can get a quick glance of the current state of the PR 
>> with low effort
>> 
>> Both of these projects are available as public repos and would love to hear 
>> your feedback. What's missing? What could be better? Jarek Potiuk and I had 
>> a discussion about maybe folding this into Apache Magpie. What do you think? 
>> https://github.com/pmcfadin/cassandra-pr-review
>> https://github.com/pmcfadin/cassandra-project-health
>> 
>> Patrick
>> 
>> 1. https://www.chaoss.community/kb/metrics-model-starter-project-health/
>> 2. https://github.com/rustyrazorblade/skills

Reply via email to