docs(intelligent-compute): document the On-Demand fallback cost limit

[pditommaso]

Summary

The Intelligent Compute page says the spotFirst provisioning model "falls back to On-Demand if Spot capacity is unavailable". That's no longer the whole story. The scheduler now falls back only while On-Demand costs at most a set multiple of the Spot price. When On-Demand is dearer than that, the task is retried on Spot and can end up failing, which the page doesn't mention. This limit already refuses fallbacks in production (seqeralabs/sched#1291).

What changes for readers

• Provisioning model row: spotFirst now says it falls back unless On-Demand is too expensive, and links to the new section.
• New section On-Demand fallback cost limit under Configuration options. It explains:

  • how the limit is measured: the highest On-Demand hourly price compared with the highest Spot hourly price, across the instance types the scheduler selected for the task;

  • the current default for each cloud provider, with a note that Seqera can change them:

    Cloud provider	Default limit
    AWS	3.25 times the Spot price
    Google Cloud	2.48 times the Spot price
    Azure	No limit

  • what happens when the limit is exceeded: the task is retried on new Spot capacity, then fails with an error message starting on-demand fallback refused;

  • how to change the limit: currently only in the pipeline's Nextflow configuration. The compute environment has no setting for it:

    • seqera.executor.machineRequirement.onDemandCostRatio sets it for all the tasks of a run;
    • the seqera/machineRequirement.onDemandCostRatio process hint overrides it for one process, the same way as provisioning;
    • not set means the default applies; 0 means no limit (always fall back, not "never"); > 0 is the limit used;

  • what to do if a pipeline must complete when Spot is scarce: use the ondemand provisioning model. The spot model never falls back.

The default values match the scheduler defaults shipped in sched 0.96.0 (sched.scheduling.on-demand-fallback-cost-ratio-{aws,gcp,azure}), with no deployment override.

Before merging

• Nextflow 26.10 is released with Add seqera.executor.machineRequirement.onDemandCostRatio option nextflow-io/nextflow#7710 (the config option and process hint). The page says "Nextflow 26.10 or later".
• seqeralabs/sched#1400 (run-level ratio) is released, and seqeralabs/sched#1491 (per-task ratio, needed for the process hint) is deployed.

Checks

• markdownlint-cli2 reports the same 48 errors before and after this change, all in existing tables. The new text adds none.

🤖 Generated with Claude Code

[netlify]

✅ Deploy Preview for seqera-docs ready!

Name	Link
🔨 Latest commit	1ceb0a6
🔍 Latest deploy log	https://app.netlify.com/projects/seqera-docs/deploys/6ac3e8f393870500083a97df
😎 Deploy Preview	https://deploy-preview-1913--seqera-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.