Migrate from Namespace to WarpBuild
Move GitHub Actions workflows from Namespace to WarpBuild. Choose runner labels, update caches and builders, test your jobs and keep a way to switch back.
Last updated:
To move from Namespace to WarpBuild, choose a runner and update any steps that depend on Namespace. After account setup and GitHub App installation, a standard job can move by changing its runs-on label. If your job uses cache volumes, Namespace actions, remote builders or private networking, you’ll need to update those too.
Start with a job you run regularly. Check that it builds or tests correctly and can reach the services it needs, then move a few more jobs at a time. This guide covers the label map, workflow changes, caches, compatibility checks, parallel running and rollback.
For platform and pricing differences, see WarpBuild vs Namespace. For a broader shortlist, see Namespace alternatives.
Prerequisites
- Connect the repository. Create a WarpBuild account and install the GitHub App with access to the repositories you plan to move. Follow the quick start, including organization approval where required. GitHub Enterprise Cloud with data residency and GitHub Enterprise Server use the Enterprise setup.
- Choose where the runners run. Use WarpBuild Cloud or configure BYOC. Linux BYOC supports AWS, GCP and Azure; Windows BYOC supports AWS and Azure. macOS runs on WarpBuild Cloud. For BYOC, configure the region, network, runner group and any custom VM image before selecting it in a workflow.
- Find the Namespace-specific settings. Search workflow files and local actions for Namespace labels, profiles, actions and environment variables. Also review reusable workflows and settings held outside the repository.
rg -n 'nscloud|namespace-profile|namespace-features|namespacelabs/|NSC_' .githubMake a list of cache paths, container registries, jobs that read saved artifacts, private services and any credentials Namespace adds to the runner. GitHub secrets and workflow permissions remain in GitHub, but the new runner still needs the right network path and authentication. Check that the new runner can assume the cloud roles it needs through GitHub OIDC. Set up a replacement for any Namespace-specific federation before removing it.
The Workflow Change
After setup, change the runner label on one job:
jobs:
build:
- runs-on: nscloud-ubuntu-24.04-amd64-8x16-with-cache
+ runs-on: warp-ubuntu-2404-x64-8xThis moves from 8 vCPU / 16 GB to 8 vCPU / 32 GB. This changes the machine running the job. You still need to set up caching and any remote builder separately.
A job using a Namespace profile also changes its runs-on value:
jobs:
integration:
- runs-on: namespace-profile-large-builder
+ runs-on: warp-ubuntu-2404-x64-16xCheck the settings behind the profile name, including its architecture, memory, image, network and cache. The label above is an example choice, not a direct translation of every profile named large-builder.
Label Map
Start with the labels below, then check that the architecture, memory, OS image and tools fit your job. Use a label with an OS version from the runner catalog so a change to the latest alias does not move your job to a different OS version.
| Namespace label | Namespace machine | WarpBuild label | WarpBuild machine |
|---|---|---|---|
nscloud-ubuntu-24.04-amd64-2x4 | 2 vCPU, 4 GB | warp-ubuntu-2404-x64-2x | 2 vCPU, 8 GB |
nscloud-ubuntu-24.04-amd64-4x8 | 4 vCPU, 8 GB | warp-ubuntu-2404-x64-4x | 4 vCPU, 16 GB |
nscloud-ubuntu-24.04-amd64-8x16 | 8 vCPU, 16 GB | warp-ubuntu-2404-x64-8x | 8 vCPU, 32 GB |
nscloud-ubuntu-24.04-amd64-16x32 | 16 vCPU, 32 GB | warp-ubuntu-2404-x64-16x | 16 vCPU, 64 GB |
nscloud-ubuntu-24.04-amd64-32x64 | 32 vCPU, 64 GB | warp-ubuntu-2404-x64-32x | 32 vCPU, 128 GB |
nscloud-ubuntu-24.04-arm64-8x16 | 8 vCPU, 16 GB | warp-ubuntu-2404-arm64-8x | 8 vCPU, 32 GB |
nscloud-ubuntu-22.04-amd64-8x16 | 8 vCPU, 16 GB | warp-ubuntu-2204-x64-8x | 8 vCPU, 32 GB |
nscloud-ubuntu-26.04-amd64-8x16 | 8 vCPU, 16 GB | warp-ubuntu-2604-x64-8x | 8 vCPU, 32 GB |
nscloud-ubuntu-20.04-amd64-8x16 | 8 vCPU, 16 GB | No Ubuntu 20.04 image | Move the job to 22.04 or newer |
nscloud-windows-2022-amd64-4x8 | 4 vCPU, 8 GB | warp-windows-2022-x64-4x | 4 vCPU, 16 GB |
nscloud-windows-2022-amd64-2x4 | 2 vCPU, 4 GB | warp-windows-2022-x64-4x | 4 vCPU, 16 GB, the smallest Windows size |
nscloud-macos-sonoma-arm64-6x14 | 6 vCPU, 14 GB | warp-macos-14-arm64-6x | 6 vCPU, 14 GB |
nscloud-macos-sequoia-arm64-6x14 | 6 vCPU, 14 GB | warp-macos-15-arm64-6x | 6 vCPU, 14 GB |
nscloud-macos-tahoe-arm64-6x14 | 6 vCPU, 14 GB | warp-macos-26-arm64-6x | 6 vCPU, 14 GB |
nscloud-macos-tahoe-arm64-12x28 | 12 vCPU, 28 GB | warp-macos-26-arm64-12x | 12 vCPU, 28 GB |
nscloud-macos-goldengate-arm64-6x14 | 6 vCPU, 14 GB | warp-macos-26-arm64-6x for compatible toolchain requirements; not a macOS 27 host replacement | 6 vCPU, 14 GB |
WarpBuild macOS 26 images include Xcode 27.0, 27.1 and 27.2 alongside the default Xcode 26.6. Select the version your build needs and check the simulator runtimes it uses. If your tests need to run on a newer macOS version, having its SDK on an older host is not enough. The 12 vCPU macOS size is available on macOS 15 and 26.
After choosing the machine, update any extra settings in the old label:
| Namespace setting | What to configure on WarpBuild |
|---|---|
-with-cache, cache tags and cache-size labels | Configure paths, keys and restores with WarpBuilds/cache@v2 on Linux or macOS. Cache storage and operations are billed separately |
| Read-only cache behavior | Use WarpBuilds/cache/restore@v2 without a save step |
-with-builders, nscloud-in-runner-builder | Create a remote Docker builder profile and select it in the build action |
| Tailscale configuration | Create the WarpBuild network configuration and use its network.name label modifier through the networking setup |
| Job priority and per-profile runner limits | Review workflow concurrency controls and the available runner capacity instead of copying Namespace scheduling labels |
| Preview image selectors | Choose the required host image and installed toolchain explicitly |
A Full Workflow
This example runs a Bun project with a committed bun.lock and a test script. Keep your project's package manager, version and commands when adapting it. The runner uses an explicit Ubuntu 24.04 label, and the cache key separates the OS, architecture and Bun version.
name: build
on:
push:
branches: [main]
pull_request:
permissions:
contents: read
jobs:
test:
runs-on: warp-ubuntu-2404-x64-8x
steps:
- uses: actions/checkout@v4
- uses: oven-sh/setup-bun@v2
id: bun
- uses: WarpBuilds/cache@v2
with:
path: ~/.bun/install/cache
key: ${{ runner.os }}-${{ runner.arch }}-bun-${{ steps.bun.outputs.bun-version }}-${{ hashFiles('bun.lock') }}
restore-keys: |
${{ runner.os }}-${{ runner.arch }}-bun-${{ steps.bun.outputs.bun-version }}-
- run: bun install --frozen-lockfile
- run: bun run testPin Bun through your project's packageManager field or the setup action's version input. Replace Namespace checkout with your supported checkout action, preserving submodule, Git LFS, fetch-depth and credential settings. Remove Namespace setup steps only after their remaining responsibilities have been replaced.
For other language ecosystems, the cache-enabled setup actions combine toolchain setup with dependency caching. To cache other directories, set the path, key and restore-keys inputs in the cache action. Check restored file ownership and permissions before relying on the cache.
Cache Migration
Namespace Cache Volumes expose persistent directories on local NVMe storage. WarpBuild's cache action restores and saves paths using workflow keys. Moving the runner label does not transfer the data or preserve volume mount behavior.
| Existing dependency | Migration step |
|---|---|
Cache Volumes and nscloud-cache-action | Identify the actual paths, then configure WarpBuilds/cache@v2 or a supported setup action. Remove assumptions about mounted volumes |
| Toolchain downloads | Keep required version pins and use the matching setup action |
| Checkout caching | Use a supported checkout action and validate credentials, submodules and fetch behavior |
| Namespace Artifacts | Move retained outputs to GitHub Actions artifacts or your object storage, and update the jobs that consume them |
| Container image and layer caching | Configure registry authentication and the chosen Docker builder separately |
| Bazel, Turborepo, Pants, Moonrepo, Gradle or sccache | Retain or replace the build-system cache endpoint. For local cache directories, configure paths and keys. Gradle workflows can use WarpBuilds/gradle-actions |
| Action downloads | Keep action references and required access in the workflow; do not carry over Namespace cache settings |
WarpBuild cloud runner caching supports Linux and macOS, not Windows. On Linux and macOS, run the job once without cached data, then again after the cache has filled. Check that both runs produce the right result. Use different cache keys for architectures or toolchains that cannot share cached files.
For remote Docker builders, create a builder profile, select it with profile-name in Warpbuilds/build-push-action, and verify target architectures, build arguments, secrets and image tags. Docker layers build up in a separate cache from your dependencies. Test a build and push to the registry you plan to use before switching the production build job.
If images or artifacts remain in Namespace storage, keep access during the trial. Move the data you need and update the jobs that read it before deleting the originals.
Compatibility Notes
| Area | What to check before rollout |
|---|---|
| Linux and Windows images | Match the OS version, architecture, memory, installed tools and shell behavior. The smallest WarpBuild Windows shape is 4 vCPU / 16 GB |
| macOS | Match Xcode, SDKs, simulators, signing credentials and host OS requirements. WarpBuild offers 6 vCPU / 14 GB and 12 vCPU / 28 GB configurations; workloads needing 56 GB require a separate sizing decision |
| Linux on Apple Silicon | WarpBuild offers 6 and 12 vCPU -xfast labels. Select these explicitly if the job depends on Apple Silicon Linux, rather than substituting a general ARM64 runner |
| Private services and egress rules | Configure the network path, allowlists and credentials. BYOC lets you apply network controls in your own cloud account |
| Concurrency | Generally available Linux and Windows runners have no plan-level cap. macOS defaults to about 100 concurrent jobs per organization, raisable through support. BYOC remains subject to cloud quotas |
| Debugging and metrics | Use SSH debugging and runner metrics correlated with job logs. Update dashboards, OTLP exports and automation that depended on Namespace tooling |
| API and MCP workflows | Map each operation to the WarpBuild API or MCP tool. Replace Namespace CLI calls and credentials individually |
| Devboxes and graphical sessions | Keep development environments and VNC or RDP dependencies separate from the runner migration. Validate the replacement workflow before removing access |
| Organization access | Configure required SSO and access policies separately from the GitHub App installation |
Run Both Platforms Side by Side
Runner selection is per job, so a repository can use both providers during the rollout.
- Pick a build or test job you run often. Note whether it passes, how long it waits to start, how long it runs and what it costs.
- Run that job on WarpBuild and leave the others on Namespace. To compare both providers, run the same commit with the same inputs in separate test jobs.
- Try a run with an empty cache and a few more after it fills. Check the test results, saved artifacts, access to private services and what happens when the job fails. Make sure deployment and publishing steps only run once.
- Check the job names used by your branch protection rules. If you gave the test jobs new names, update the required checks when you’re ready to switch.
- Once the job works as expected and you’re happy with its speed, move a few more jobs at a time.
Keep the old workflow settings and Namespace resources while you test. Before moving release jobs, also check signing, access to production services and any jobs that read their artifacts.
Billing During the Parallel Run
Moving jobs does not cancel the Namespace plan. The Developer plan is pay as you go. The Team and Business plans have monthly commitments and included usage. You still pay for any remaining usage, stored data and plan commitments while you try WarpBuild.
WarpBuild runner usage is billed per minute by shape with no base subscription, platform or seat fee. Include cache storage and operations, snapshots, remote builders and any other services used in the trial. For BYOC, include the runner management charge and your cloud infrastructure costs. See WarpBuild pricing.
Compare the full cost of running the workflow, both with and without cached data. Set a budget and spend alerts before moving more jobs. WarpBuild hard caps leave new jobs queued when the cap is reached.
Rollback
Revert the migration pull request to restore the previous labels and actions:
jobs:
build:
- runs-on: warp-ubuntu-2404-x64-8x
+ runs-on: nscloud-ubuntu-24.04-amd64-8x16-with-cacheThe label change is only part of the revert if you also changed checkout, caching, builders or artifact references. Also change back any settings you updated outside the workflow files.
Before rerunning the workflow, confirm that:
- The Namespace GitHub App still has repository access and the required profiles and builders remain available.
- Credentials, federation, network rules and registry access still work.
- The artifacts and images the job needs are still there. Reverting a pull request will not bring back deleted data.
- Cache entries are still available. Retained Namespace caches may be reusable; expired or removed entries require a cold run.
Run the job with its old settings and check the results before switching the other jobs back.
After the Move
Once the jobs you want to move are working on WarpBuild, check scheduled workflows, reusable workflows, release jobs and anything that reads their artifacts for leftover Namespace settings. Finish moving any images and artifacts you need before deleting their original storage.
Then remove unused Namespace App access, tokens and federation settings, delete resources you no longer need, and check the terms for cancelling or downgrading your plan. Keep any access still needed by Devboxes or other workloads that did not move.
For workflows with expensive setup, snapshot runners can reuse a prepared environment after the basic migration is stable. They support WarpBuild Cloud Ubuntu runners, exclude BYOC, Windows and macOS, and expire after 15 days. Try snapshots and other build improvements one at a time so you can see what each change saves.
FAQ
How much of my workflow changes?
The runner label changes for each job you move. Namespace-specific checkout, cache, builder, artifact and registry integrations need separate updates. Standard workflow steps can stay in place after you check image, tooling and access requirements.
Do my caches transfer?
Caches do not transfer automatically. Set up the paths and keys in WarpBuilds/cache@v2 on Linux or macOS, then run the job a few times to fill the cache. WarpBuild cloud runner caching does not support Windows. Retained Namespace caches may still be available if you roll back.
Can I run both providers at once?
Yes. Runner selection is per job, so some jobs can use WarpBuild while others remain on Namespace. Use separate jobs or trial workflows to compare providers, and keep deployment or publishing steps from executing twice.
What happens to my Namespace plan while I test?
Moving jobs does not cancel your Namespace plan or remove stored data. Existing commitments and remaining usage continue to apply. WarpBuild runner usage has no base, platform or seat fee, with cache, snapshots, builders and BYOC infrastructure contributing separately where used.
How do I roll back?
Revert the migration pull request and restore any external settings changed during the trial. The previous GitHub App access, credentials, profiles, builders and stored data must still be available. A Git revert cannot recreate deleted resources.
Does anything change for secrets and permissions?
Repository and environment secrets and workflow permissions remain configured in GitHub. Test that the new runner can use them and reach the required services. Namespace-specific federation, injected credentials, IP allowlists and private network access need separate replacement or configuration.
Start with $10 in free credits
Change the runner label in your workflow and keep the rest of your GitHub Actions setup. Runner time is billed per minute.