A classic productivity killer is to get stuck on something and keep banging at the wall over and over. Every engineer has encountered this. The usual advice for this is to reach out for help when you’re stuck.
The problem with that advice is that it’s not always applicable, you could be diving into a rabbit hole for hours and still get the feeling that you’re not stuck, but instead jumping from one hypothesis to another. You are technically, not stuck, so why ask for help? And the rabbit hole exploration continues…
So instead, let’s redefine the criteria for asking for help: if you do not make tangible progress toward your goal for a few hours (usually, if more than two) it’s time to ask for help. Asking for help, even if there are other roads to explore, is not a failure. It’s instead about actively seeking to maximize your impact. A fresh pair of eyes, someone with a different context, etc … can help you to see if it’s time to change directions, go around the problem, or perhaps simply solve the problem.
Our team scope is very large, and it encompasses a broad range of domains and different skillsets, so this particular aspect of engineering is magnified, making it even more important to ask for help at the right time.
So, if you don’t see tangible signs of progress after a two hours on something, reach out to your peers! We’re a team and that’s exactly why it’s here for, we win together, and together we win even more.
- Identify the problem, who has this problem (the customers), and potential solutions.
- Execute on a solution, in line with our principles, and communicate progress back within the team (e.g. via the planned work process or unplanned work process)
- Roll out the change to our customers.
- Artifact: Sharing it in our channels (Slack and newsletter). Educating our customers about the change, and helping them adopt it.
- Tracking adoption and feedback.
- Artifact: feedback or metrics on adoption.
When planning work, the relevant work lead should:
- Work with the PM/EM to create the roadmap tracking issue using the tracking issue bot. A loose scratch doc can be created in the team Drive as well, but ultimately the tracking issue is the concrete artifact we share with stakeholders and the company. The tracking issue description should have the following:
- Problem: Articulate the problem space and requirements.
- Scope: Scope the work so that it can be completed in a single bet cycle, identifying:
- Boundaries: Define what work is within the scope and what work is explicitly out of scope. (Out-of-scope boundaries, or “non-goals”, are worth writing down and they help prevent scope creep.)
- Definition of done: the desired end state that needs to be achieved for a work item to be considered “done”, in addition to:
- a rollout strategy
- “non-functional requirements” such as testing, monitoring, and documentation
- a retrospective (e.g. by consulting relevant stakeholders) to identify if goals were achieved and follow-up work needs to be done
- Create a implementation plan by creating and tagging GitHub issues to be tracked in the roadmap issue.
- Issues should be labeled with days estimates (as supported by the tracking-issue-bot), other metadata where helpful, a rough order of operations, and outline of what can be parallelized.
- Dependencies on other work are identified and tracked.
- Maintain an understanding of the progress of the work.
- Have a clear picture of what’s being executed, what’s coming up, and overall timelines.
- Ensure required tasks are delegated.
- Continually evaluate whether our priorities are right.
- Work with EM/PM if any adjustments are needed.
- Create follow-up issues for any known or outstanding bugs.
Note that teammates can also support a work lead on all of the above - it is not exclusively the responsibility of work leads.
We aim to allow teammates the flexibility to work on incoming requests, tackle proactive improvements, and invest in long-term efforts on our roadmap, so as a rule of thumb:
- We aim to spend 20% to 30% (~2-3 days every 2 weeks) of our time on making proactive impact, i.e. working on things that are aligned with the team’s mission, but aren’t on our roadmap.
- If over 50% (~5 days every 2 weeks) of our time is spent outside of planned work (i.e support requests), we opt to discuss the scope and priority of the work with the team first.
We encourage you to log unplanned work in GitHub issues, the DevInfra-scratch repository, or in the daily Geekbot updates.
If the work is a response to an active problem or urgent request that must be prioritized outside of our planned work, create an issue and add the
reactive label so that we can prioritize the work and review our reactive work load at the end of a bet period.
Note that even unplanned work should take into account a rollout strategy, especially if it is implementing improvements based on feedback.
For each work item, we ask the following questions:
- Who needs to know about this change?
- How and where are we reaching them?
- What do they need in order to adopt this change?
From the answers, we extend the Definition of Done for the item to include these steps:
- Create an announement about the change in immediate channels (relevant Slack channels)
- Add an entry to the Developer Infrastructure newsletter
- Update documentation if applicable
- Create resources to help engineers adopt the change
- Create content to educate engineers about the background and reason for the change
Remember that pull requests are simply a dialogue between team members about changes that we’re shipping as a team and we’re here to help each others to achieve it.
On top of the company wide process for reviewing pull requests, we follow these principles when it comes to review each others:
- We own our PRs:
- Ex: If looking to be reviewed by someone in particular, just explicty add them in the reviewers (useful to get a thorough review when a teammate has a lot of context for example).
- Ex: Reach out for others if a review is urgent.
- Ex: PR author is responsible of taking the decision on when it’s ok to merge it, to keep adding changes or to add those in a different PR.
- We eagerly communicate in our PRs:
- Ex: If a review was expedited to unblock the PR, just say so and let its author bear the responsibility of merging or not their PR.
- Ex: If only a subset of the PR content is familiar to you, you can still review, just say what you didn’t cover.
- Ex: If I don’t understand why something is done in a certain way, just ask. There are no dumb questions.
- Ex: If you’re happy to see a change be merged, just say it! Async doesn’t mean we can’t be supportive of each others like we’d have done if we’d be sitting next to each other.
- When in doubt, act!
- Ex: If you’re not sure that a piece of code is going to create a problem, just say so, that’s always better than not posting anything and you can always come back later and confirm.
- Ex: A PR should fix a problem, but that’s not entirely sure. If merging tells you the answer, and that’s shorter than verifying, just do it unless the consequences are too high.
- Ex: A PR has been merged? You can still review it, its author will open a new PR if you spot a mistake!
As of , we use GitHub issues and projects to keep track of our work.
For a lower level view of our day-to-day progress, we use the Developer Infrastructure GitHub project. This board automatically imports issues with the
team/dev-infra labels. Additional views are avaialble on the board for specific purposes, such as viewing active roadmap issues.
Roadmap tracking issues are what we communicate to teammates outside of the Developer Infrastructure team what we are working on and what we plan on working on. Roadmap issues should be scoped to a bet period, and be a tracking issue that encapsulates everything outlined in the planned work process.
They should have the
roadmap label, one (or more) appropriate
dev-infra/... label(s), as well as the following text in the description to help the tracking issue bot find relevant issues:
<!-- OPTIONAL LABEL: roadmap -->
A project board view for roadmap tracking issues is also available here.
We collect meeting notes and planning artifacts in a shared Google Drive folder.
We encourage the liberal use of “scratch documents” - temporary, short-lived Google Docs that are used as discussion forums and a dump for ideas. Concrete, permanent artefacts should be created from them, e.g. via the planned work process.
If you have a question that relates to our areas of ownership, the #discuss-dev-infra channel is the best way to reach us. Tag the
@dev-infra-support handle if something is urgent (think of this as pulling the fire alarm) and requires our immediate attention. Our on-call teammate will jump in.
If you think Developer Infrastructure is the right team to address a bug or other request, please create an issue. The issue should include:
- A short description of the ask
- A more detailed explanation of the background, context, and problem that needs solving
- Any guidance related to the impact of this issue
- Any extra information that will help us solve or prioritize the issue
- The label
This section documents tips for teammates on support, which is handled via our support rotation. General tips:
- We are not responsible for every issue in CI - see build pipeline support and CI support responsibilities.
git rebaseis very easy for teammates unfamiliar with
gitto do incorrect - if a branch update is required, err towards recommending a
git pull origin main).
- Be wary of issues achieving joke/meme status: it often means that it is actually quite a disruptive issue, and warrants a closer look.
- Be liberal in asking for a video or screen recording to demonstrate DX issues: this can help us understand and emphathise with DX pains.
Tips for tracking support work:
- Make sure to update DevInfra-scratch for any notable support efforts and problems you ran into.
- If a sizeable amount of work is surfaced (e.g. an active problem or urgent request), follow the unplanned work tracking guidance to help us keep track of how much reactive work we are tackling, and to ensure the work gets prioritized as needed.
Build pipeline support pertains to our continuous integration. The goal of this process is to identify the right person to drive a fix on an issue, rather than actively fixing every issue that arises.
The on-call support teammate should monitor the pipeline through channels like #buildkite-main for flakes and notifications from
Monitoring generally means taking a look at a random sample of 2-3 failures in #buildkite-main throughout the day and identifying if it might be a legitimate issue, of if it is a flake - refer to the flakes playbook to learn more.
If there are any issues, ensure issues are followed up on:
- Infer the owner based on the contents of the issue *e.g. through product names, git history, and/or other context) and reach out for assistance:
- If a team can be inferred, ping the
@$TEAM-supporthandle in Slack for assistance, escalating to
@$TEAMif no support handle or teammate is available.
- If no team is easily inferred, ping the most recent author via
git blamewhere relevant for assistance.
- If a team can be inferred, ping the
- Guide the teammate towards a resolution for the issue by following our broken builds process (also see Continuous integration: Flakes).
The DevInfra team is not responsible for all the tools and tests that run in Sourcegraph’s CI infrastructure.
If a tool or test is behaving in an unstable manner, the team using the tool or test (see build pipeline support for how we infer ownership) has the responsibility of leading an investigation into what might be causing said instability, with the assistance of the DevInfra team if helpful.
The team leading the investigation should either fix the issue directly, or if the issue requires changes in the DevInfra team’s ownership/responsibility areas (e.g. increasing resource limits for build agents, or making significant changes to the pipeline generator), request the desired changes through an issue tagged
For a higher-level understanding of our responsibilities, see our guiding principles.
We use OpsGenie to manage a support rotation that changes weekly. The person on-call will be pinged when someone tags
@dev-infra-support in Slack. That person is expected to prioritize responding to the support request, but anyone who sees a request can and should jump in.
For each major body of work, we assign an engineer to be a work lead. Work leads are responsible for working with the PM/EM to articulate the problem space, requirements, proposed solution, results, and measures of success. They also scope a delivery plan and create issues to be tracked in the tracking issue. The work lead helps answer questions about priority order, scope change, etc., and provides updates at team syncs. They are not solely responsible for executing on the work, which is a team effort.
See the planned work process for additional details.
- Meeting leaders rotate weekly, and the leader should prepare the agenda at least 24 hours ahead of time.
- All teammates should add items to the agenda before the meeting begins.
- Any action items that come out of meetings should be considered a high priority.
We connect weekly to discuss any topics that are best addressed in person. Our notes are available here.
We connect bi-weekly to celebrate wins from the past two weeks, give cheers to our teammates, and reflect on what we’ve liked, not liked, and learned for going forward. The focus of the agenda for this meeting is team health, as opposed to day-to-day work.
Every week we hold an informal, totally optional coffee break to connect socially and hang out together as humans.
There is a weekly
sg hack hour that Thorsten Ball and the DevInfra team co-host weekly Fridays from to for anyone interested in making contributions to the Sourcegraph developer tool.
When the hack hour starts, a meeting link will be posted in #dev-experience.
To learn more about contributing to
sg, check out the contribution guide!
At the start of each week, the TL posts a reminder in
#team-dev-infra indicating who is on support and who is leading meetings.
All engineers update the Developer Infrastructure project board in advance of the weekly sync.
The status field for each issue in the project board roughly represents the following:
- Unplanned: No concrete plans to tackle yet
- Next iteration: Something to consider for next iteration
- Planned: On somebody’s immediate plans (i.e. will be tackled roughly within this iteration)
- In progress: Actively being worked on
- Blocked: No progress can be made due to a dependency
- External: DX-related work being handled by other teams
- Done: That’s a wrap!
The PM and work leads own the roadmap issues. They solicit input from the team to make sure the problem space is thorough, measures of success are ambitious-but-reasonable, and technical details are correct.
The PM updates this page at the start of each month.
The DevInfra team is responsible for a monthly (ish) newsletter to highlight developer experience updates (not just those lead by the team). Learn more about it and see previous issues in the newsletter archive.
To prepare a new issue of the newsletter, create a PR for the latest newsletter issue here following the conventions in the previous newsletters. Some tips:
- You can refer to
dx-announceissues and PRs for content ideas!
- Adding a
closed:>YYYY-MM-DDwill filter the list down to just things that have been closed since the last newsletter issue.
- Adding a
- To include images, either follow the official guide or upload images to a GitHub issue - this will provide a shareable link.
Once the newsletter is ready and reviewed, merge the PR. Then copy and paste the rendered newsletter from the handbook (you can set this up locally with
yarn dev) into a draft newsletter. You will need to remove the background color from the pasted content, but the formatting should otherwise just work.
Verify the output looks good, and email it to firstname.lastname@example.org.
We maintain a repository with experiments and notes from the DevInfra team,
sourcegraph/dev-infra-scratch, in an effort to improve knowledge-sharing and log our explorations and trends.
We encourage you to log anything you want in this repository, particularly insights from pairing sessions and so on!