A little over a month ago, the Concourse team sent a survey out to the community. The purpose of this survey was to gain insight into our users as well as measure our year-over-year growth. In the process of learning about how you all deploy and manage Concourse, we also received tons of great feedback about what’s working well and what needs work in order to make Concourse even better. We’re excited to share our findings!
A huge thank you to everyone who responded. At the time of this writing, we’ve received over 100 responses and that number is still climbing. Your contributions are valuable, and learning about how different segments of our user base works with our product is going to help us make Concourse even better in 2020!
Feature requests, areas to improve
This is a big one. The community wants more control for administrators and operators, more options for integrations, and more power over resource types configuration. We also learned a lot about the specific ways Concourse is making life more difficult than it needs to be in terms of configuring tasks, pipelines, teams, and the product itself.
Code and configuration duplication is a serious issue, and our users want more powerful templating tools to help them split pipeline configuration into more manageable chunks that will be easier to reason about and maintain.
We’re also paying particular attention to the number of responses that were focused on git integration and GitOps workflows. If you have a way of using Concourse that you feel isn’t well represented by the current featureset or CLI/UI, please @mention us on Twitter or drop by Discord and tell us about it.
Web UI 🖥
Concourse’s web UI is a hot topic! While most of the web UI feedback is positive, there are lots of suggestions on how to improve it or what to add next. Feedback from the survey about the web UI could make for its own blog post, so in the interest of being brief, I’m just going to touch lightly on the strongest signals/insights that were generated.
A number of respondents called to attention the ease of use and clarity of information in the current UI. While we’ve been continuously iterating to add text labels instead of just icons where possible, and to add clarifying tooltips elsewhere, there’s clearly a need for more. In addition to several smaller tweaks, we have work underway around adding the minimum viable Favorite Pipelines functionality that will be built upon to extend the Archiving Pipelines functionality introduced in v6.1.0 to the front end. Hopefully these fixes will make a big impact, decluttering Concourse dashboards and making it a lot faster to find what you want at the same time.
Another area where we can clearly improve is by adding more detail to the dashboard. Users are requesting more options for adding notes, tracking an audit trail of actions in the UI, clearer and more detailed error messages, and more statistical information like build duration and lead times. We’ll be looking at the possibilities in this space over the coming months. If you have ideas, start a Discussion on Github.
Runtime Improvements 📈
In addition to more stability and performance, the community puts a high level of importance and value on improving the efficient use of check containers, global locks on resource checking, and the ability to clear cached resource versions of a worker on demand with fly.
Docker Enhancements and Performance 💨
We hear you. 😀
Comments from the community emphasized optimizing docker-image resources, facilitating docker in worker containers, and better reporting on docker image status. There are a number of different voices in this conversation all with very different strategies for how they use Concourse, and we’re sorting through feedback to help us prioritize low hanging fruit and high value enhancements that the team can prioritize.
Additionally, we’re actively monitoring issues and continuously collecting data on Docker performance so that we can make more improvements - we understand that every last bit of performance we can squeeze out of Docker interactions results in a huge benefit to many of our users.
Stability, Kubernetes, Documentation ⚖️ 🚢 📚
These issues have remained top of mind in the community for the past few years, and this year’s survey is no exception. From a stability perspective, the team has made great strides with the release of the new algorithm in version 6.0.0. The team has also taken further steps into being more k8s native by beginning an ongoing track of work dedicated to running K8s workloads. And lastly, our documentation work is ongoing - we hope to prioritize more ‘getting started’ materials for beginners in order to enable new users to climb the learning curve faster than before. For more advanced users, we also plan more documentation around topics like autoscaling, tracing, and build statistics, among others.
How long have you used Concourse?
Most of the people who responded indicated they had been using Concourse for one year or less. It’s great to see that more people are picking up and experimenting with Concourse with each new release, and it’s just as exciting to see that people stick around: more than 45% of respondents said they have been using Concourse for 2+ years. Whenever we interpret feedback from the community, we want to make sure we’re taking into account the experiences of both newcomers, established users, and very experienced power users. Each segment experiences different challenges, and prioritizes different parts of the product.
Other CI CD tools used
Another dimension that’s helpful to understand is the related experiences that each survey respondent is equipped with. When looking at other CI/CD tools that our community employs, Jenkins is still the top dog, accounting for nearly 30% of the tools mentioned. Github Actions has seen a rise in adoption since its initial release, and Travis, Gitlab, Bitbucket, and CircleCI are all fairly common options as well.
How did you find out about Concourse?
Pivotal Software (now VMware) has been Concourse’s largest supporter since the project’s inception. In previous years, it was common to see more than half of respondents say they were introduced to Concourse CI through a Pivotal Labs engagement, or through Concourse’s role in automation of the Pivotal Platform, Pivotal Cloud Foundry. Now the community has started to branch out, with only 22% of people reporting that they learned about the product through Pivotal.
The majority of users seem to have found Concourse organically, through search engines or social media. We’re hoping to expand the use of our blog this year to help support the number of people hunting for Concourse content. Be on the lookout for more tutorials, advanced operations articles, and general updates about the Concourse product development and roadmap.
We’d love to grow that Conference or Meetup section in 2020 - who’s up for a remote meetup over Zoom? 🙌
Why use Concourse?
When asked about the very important _ why _ behind their Concourse usage, concerns about Open Source tooling and flexibility were top of mind. The special emphasis that Concourse put on reproducibility and user interface also ranked highly, along with Concourse’s scalability and overall feature set. Scalability is always a huge concern for the team, as we see enterprise customers frequently testing the limits of their tooling (sometimes with hundreds of Concourse clusters, many thousands of teams, and many hundreds of thousands of pipelines). Likewise, reproducibility is a commitment we’re not planning on straying from any time soon.
We released the survey just as v6.0.0 of Concourse was being finalized, so it was only close to the end that we started to see people upgrading to v6. We’re thrilled, nonetheless, to see so many people had already upgraded to v5.8.x. Together, versions v5.8.x and v5.7.x represented the majority of survey respondents, with a low (>10) rate of responses for any other version.
To those 12 users who are still on v4.x.x and 7 users still on v3.x.x, feel free toget in touch on the Concourse Discord if you need any help upgrading! You can find all of the wonderful reasons to upgrade in therelease notes, and we’ll write blog articles in the coming months highlighting some of the latest and greatest new features and optimizations, as well as some upcoming enhancements on our roadmap.
The data gathered shows that the majority of respondents are working with Concourses organized with fewer teams. And when it comes to users…
… we see a lot of smaller Concourse instances of under 10 users. There are also a few examples of large, enterprise scale deployments of 100+ users over 50+ teams. On the Concourse team, we frequently reach out to enterprise customers for special feedback on more massive implementation concerns. We also survey and interview members of the open source community to make sure we’re building solutions that scale down to single users and small teams.
If you’d like to add your voice, feel free to join in on theConcourse Discussions board.
Docker remains the most frequently used deployment method, but the margins are slowly shrinking, and there’s more even distribution across other popular options than we’ve seen in past years.
Nearly identical numbers of responses came in citing Kubernetes (via the Helm chart), BOSH , and VM deployment strategies, reinforcing both our interest in facilitating K8s workflows and supporting our substantial BOSH user base.
Concourse Usage Style
This year we asked about our users’ usage style - specifically, what sort of development scenarios they were using Concourse to facilitate. Concourse remains an Infrastructure Automation powerhouse, and a similar number of users are using it to perform CI for web development and deploying software as part of their path to production.
Linux workloads represent the vast majority for the Concourse community. We’re also paying attention to special concerns for those running Windows and Darwin workloads, however this knowledge will help us prioritize fixes to help the largest group of users possible.
Finally, when asked about their preferred IAAS, AWS takes the top position again for the third year in a row. We consistently see a strong vSphere presence from enterprise customers, but it’s really interesting to see the variety of setups that the open source community as a whole employs when deploying Concourse.
A recurring topic that comes up in conversations with customers and internal teams at VMware is the sheer variety of ways that Concourse can be set up and put to work. Running this survey further reinforces that idea, giving us insight into an even larger number of configurations and implementations than what we see during our day to day enterprise development and support.
It’s also interesting to reflect on how far the project has come since Concourse CI was first introduced. We’re nearing 22k commits from over 318 contributors adding up to 333 releases as of the typing of this sentence, and we’re looking forward to speeding up even further in 2020.
Of course, we want to make sure that we’re developing the right features, prioritizing the right fixes and enhancements, and validating that each step we take has been made in the right direction. In the upcoming months we’ll be consolidating all of these ideas into a new high level roadmap that sets out quarterly milestones for the team.