Can Your Daytona Deployment Drain an Outage Backlog?
Compare archived Daytona v0.190.0 with Daytona BYOC using a seeded batch. Measure submission recovery and the rate of verified completions while new jobs keep arriving.
Use a batch that exposes partial recovery
The archived v0.190.0 Compose excerpt shows an API configured with database, Redis, and OIDC settings. A responsive API can therefore be an early milestone while authentication or job execution still needs verification. In a disposable deployment, submit a batch of numbered jobs that each write their ID and a fixed checksum to the test harness. Capture accepted IDs separately from verified results. During an operator-controlled restart of the disposable installation, keep submitting from the harness at a fixed rate, then record when new submissions are accepted and when verified completions resume. Compare the final manifest with the original set of IDs. A missing ID exposes a gap even if the service looks healthy, and a repeated ID exposes replay that your harness must deduplicate. Preserve the client version and resolved service image identifiers so another operator can repeat the drill against the same build.
daytona/docker/docker-compose.yaml at v0.190.0 · daytonaio/daytona · GitHubMeasure what runner ownership can fix
Daytona BYOC pairs your runner infrastructure with Daytona's control plane. In a custom test region, repeat the same batch and measure the runner recovery path with the control plane available. Log separate timestamps for runner return, the first successful new submission, and the first verified result. Use the full interval through verified output when estimating how long work stopped. With twelve new jobs per minute and eighteen verified completions per minute after recovery, a ten-minute interruption adds 120 jobs and takes twenty more minutes to clear, even starting with an empty queue. If successful throughput falls to twelve, the accumulated backlog persists while arrivals continue. Keep the control-plane dependency visible in your decision: a runner recovery drill qualifies that failure path, while the behavior during a control-plane outage needs its own evidence.
Bring Your Own Compute (BYOC) | DaytonaDefine which component you need to own
Daytona’s public v0.190.0 release is available in an archived, read-only repository. It gives you a specific baseline to evaluate. If policy requires your team to operate the API, scheduler, storage, and execution hosts, identify who would own fixes and recovery before choosing that baseline or a fork. A fork is a candidate once you can identify its maintainers, patch process, and compatible client releases.
If the requirement is local execution alone, include Daytona Bring Your Own Compute in the evaluation. If shared vendor infrastructure is acceptable, include Daytona’s built-in regions. Give each candidate the same job and acceptance criteria. Check where the code runs, which services receive metadata and artifacts, and who restores service when a runner or the control plane fails.
Release v0.190.0 · daytonaio/daytona · GitHubRebuild from an inventory you can review
Open the v0.190.0 Compose definition and inventory the service images, published ports, credentials, and persistent storage. The visible file includes an API image reference with no tag and an API port mapping. Record resolved image digests and the repository revision before starting the stack. Review settings and use test credentials on a private evaluation host with disposable inputs.
Once the services are up, use the client version you intend to ship to create a sandbox, run a benign command, retrieve its output, and delete the sandbox. Repeat from a freshly prepared host using the same inventory. Compare the resolved images and lifecycle results. If a clean host gets different images or the client fails at a lifecycle step, fix the version inventory or confirm client compatibility before spending time tuning the runners.
daytona/docker/docker-compose.yaml at v0.190.0 · daytonaio/daytona · GitHubTest runner ownership with a failed job
Daytona Bring Your Own Compute attaches runner nodes to custom regions while Daytona manages the sandboxes through its control plane. If execution locality is the reason to consider it, ask which data passes through the control plane and what changes when the region has its own proxy or snapshot manager. Verify those answers against your written policy for repository contents, job metadata, and artifacts.
In an illustrative pilot, submit two jobs, each writing a distinct marker, and take one designated test runner offline before its job completes. Record what the client reports, whether the second job runs, and whether the interrupted job can restart without duplicating its output. Repeat against the archived stack if that is a viable option. Compare recovery time and the work required from your on-call team, then check which operations the service agreement assigns to Daytona.
Bring Your Own Compute (BYOC) | DaytonaPrice the maintenance you would take on
Daytona documents built-in shared regions in the US and EU where it handles provisioning and maintenance. Compare that option with Bring Your Own Compute and the archived stack using the same workload, region, and recovery target. Keep a cost line for work your team owns, including rebuilding images, patching the stack, and restoring storage. If teams staff these responsibilities differently, estimate each line separately.
For an illustrative monthly model, assume $120 for hosts, $30 for storage and backups, and four maintenance hours at an internal rate of $75 per hour. The archived deployment would cost $450 on those assumptions, before any additional egress or on-call expense. For the managed options, verify the applicable compute, storage, and transfer units and ask whether a custom region adds fees or minimum commitments. Include your own runner bill for Bring Your Own Compute.
Bring Your Own Compute (BYOC) | Daytona