Add seqera.executor.machineRequirement.onDemandCostRatio option

[pditommaso]

Summary

Adds support for the Seqera executor's cost limit on the spot → on-demand fallback. It can be set for the whole run in config, or per process with a hint, just like provisioning. The scheduler side is seqeralabs/sched#1400 (run level) and seqeralabs/sched#1491 (per task).

With provisioning = 'spotFirst', the scheduler falls back to on-demand when spot capacity runs out, but only while the highest on-demand hourly price of the instance types it selected is at most a set multiple of their highest spot price. Otherwise it retries the task on spot, and the task fails if spot capacity stays unavailable, with an error starting on-demand fallback refused. The scheduler applies a default per cloud provider (currently AWS 3.25, Google Cloud 2.48, Azure no limit).

The config value is the default for every task of the run:

seqera {
    executor {
        machineRequirement {
            provisioning = 'spotFirst'
            onDemandCostRatio = 2.5
        }
    }
}

A process hint overrides it for that process:

process align {
    hints 'seqera/machineRequirement.onDemandCostRatio': 0
    ...
}

onDemandCostRatio	Behaviour
not set	the scheduler's default for the cloud provider applies
0	no limit: tasks always fall back to on-demand, whatever the price. It does not disable the fallback; use provisioning = 'spot' for that
> 0	this limit, used instead of the default

Tasks with different limits run on separate clusters. Tasks that set a limit, 0 included, don't use the compute environment's warm pool. For now this is the only way to set the ratio: Seqera Platform has no compute-environment setting for it.

Changes

• MachineRequirementOpts: new onDemandCostRatio (Double) config option. As a @ConfigOption field it is automatically a supported machineRequirement.* hint, merged over the config value like the other fields.
• SchemaMapperUtil: sends the ratio on the run's machine requirement (the run default) and on each task's (config value or hint). A shared toOnDemandCostRatio helper checks for null rather than truthiness, so an explicit 0 is sent and not dropped, and a negative value fails with a clear error.
• Docs: a new entry in reference/config/seqera.mdx. In executor/seqera.mdx, the hint is added to the supported list, plus an "On-demand fallback cost limit" section with the config and hint examples, the unset/0/> 0 table and the cluster/warm-pool notes.
• Dependencies: sched-client 0.79.3 → 0.96.0, and the shared libHttpxVersion 2.7.0 → 2.7.1 to stay in step with the version sched-client is built against (2.7.1 is a bug-fix release: token-refresh client leak).

Before merging

• seqeralabs/sched#1491 merged (29 Sep). The per-process hint takes effect once that is deployed; the config option already works. Dev deploys from master; prod needs the next sched release.
• Replace sched-client:0.96.0-SNAPSHOT with the first sched release that includes seqeralabs/sched#1400. The per-task change needs no new client, because the field is already on the shared MachineRequirement type.

Tests

• MapperUtilTest: unset / 0 / 2.5 are mapped correctly for both the run and the task requirement, and a negative value is rejected.
• HintHelperTest: a hint of 0 overrides a config value of 3.25 (0 is a real value, not "unset").
• nf-seqera: all 260 tests pass against the published sched-client:0.96.0-SNAPSHOT (0.96.0-20260929.120116-9), which includes seqeralabs/sched#1400 and Limit on run name length #1491. With the sched-client 0.96.0 release and lib-httpx 2.7.1: nf-seqera 255, nf-tower 430 and nf-commons proxy tests pass.

Related: seqeralabs/sched#1400, seqeralabs/sched#1491, seqeralabs/docs#1913

🤖 Generated with Claude Code

[netlify]

✅ Deploy Preview for nextflow-docs ready!

Name	Link
🔨 Latest commit	21ea2bd
🔍 Latest deploy log	https://app.netlify.com/projects/nextflow-docs/deploys/6abd4f3e0162d100088245e2
😎 Deploy Preview	https://deploy-preview-7710--nextflow-docs.netlify.app
📱 Preview on mobile	Toggle QR Code... Use your smartphone camera to open QR code link.

---

To edit notification comments on pull requests, go to your Netlify project configuration.