Cloudflare Workflows cuts default Paid retention to 7 days: configure before post-mortems break
Cloudflare reduced the default instance retention for new Workflows on Workers Paid from 30 days to seven days; review new pipeline configurations, decouple success from error retention, and prevent silent post-incident debug gaps.
Decision: If your team deploys new Cloudflare Workflows on Workers Paid, check your instance retention settings immediately. Cloudflare reduced the default retention period for completed and errored workflow instance state from 30 days to seven days for all workflows created on or after September 10, 2026. Workflows created before this date keep their original 30-day retention, but newly deployed pipelines will automatically purge execution history after one week unless code or dashboard options override the default. 12
The quick read
| Aspect | Previous behavior | Current behavior (from September 10, 2026) | Operational action |
|---|---|---|---|
| New Workflows on Workers Paid | 30 days retention for completed and errored instances | 7 days default retention for completed and errored instances | Set explicit retention parameters in code or dashboard if post-incident audits require longer windows. 1 |
| Existing Workflows | 30 days retention | Unchanged at 30 days retention | Review legacy workflows before recreating them, as any newly created workflow receives the seven-day default. 1 |
| Workers Free plan | 3 days default and limit | 3 days default and limit | Continue designing workflows around a 72-hour operational review window. 12 |
| Retention ceiling | 30 days maximum | 30 days maximum on Workers Paid | Use explicit configuration to retain up to the full 30-day ceiling whenever compliance or triage requires it. 12 |
| Granular controls | Shared retention window | Separate successRetention and errorRetention properties | Adopt asymmetric retention to retain errors for 30 days while purging successful executions quickly. 13 |
| Storage billing | Billed on peak retained state at $0.20/GB-month | Billed on peak retained state at $0.20/GB-month | High-frequency pipelines see an immediate steady-state storage drop under the new default. 4 |
What changed at the boundary
Cloudflare Workflows run multi-step code on Cloudflare Workers, automatically persisting step outputs and execution logs so steps can sleep, retry, and resume reliably. Cloudflare stores this instance state after a workflow run finishes, allowing operators to inspect past step outcomes, examine return payloads, and trigger replays. 23
Before September 10, 2026, Workers Paid retained completed and errored workflow instance state for 30 days by default. Under Cloudflare's new policy, any Workflow created on or after September 10, 2026, defaults to seven days of retention for both successful and errored runs. 1
Existing Workflows are grandfathered under their previous retention behavior. Cloudflare leaves existing pipelines at 30 days. The Workers Free tier also remains unchanged, maintaining both a default and a hard limit of three days. Across all plans, the maximum retention limit on Workers Paid remains 30 days. 12
The debugging and billing trade-off
Cloudflare introduced the seven-day default to reduce storage costs automatically. Workflows billing calculates storage using gigabyte-months, billed at $0.20 per GB-month beyond the 1 GB-month included on Workers Paid. Cloudflare measures this storage across all instances, including running, sleeping, errored, and completed executions. 4
For high-throughput pipelines that process thousands of instances daily, holding 30 days of completed step payloads creates a steady-state storage bill that grows far larger than active execution needs. Cutting the default retention to seven days reduces that accumulated baseline by more than 75%. 14
The trade-off falls squarely on post-incident investigations. Engineering teams operating bi-weekly sprints or monthly audit cycles often inspect failed workflows weeks after execution. If an incident occurred eight days ago, a newly created workflow will have already purged the instance state, intermediate step return values, and failure stack traces. Teams relying on past instance logs for monthly financial reconciliations, customer support triage, or compliance verification face silent data loss if they assume the platform retains a month of history. 1
Configuring asymmetric retention
Cloudflare provides programmatic and dashboard controls to avoid this debugging gap without sacrificing storage efficiency. In the Workers API, the
MY_WORKFLOW.create() method accepts an optional retention configuration containing two separate duration fields: successRetention and errorRetention. 13This separation enables asymmetric retention policies:
// Define instance retention when triggering a workflow run
const instance = await env.MY_WORKFLOW.create({
params: { customerId: "cust_9812", batchId: "b_2026_09" },
retention: {
successRetention: "2 days",
errorRetention: "30 days",
},
});Under this pattern, successful executions purge after two days, keeping storage costs minimal for standard traffic. If a step throws an unhandled error, the instance state and step context remain pinned for the maximum 30 days, giving engineers the full investigation window. 13
Teams can also configure retention periods per Workflow and per instance directly in the Cloudflare dashboard under the Workers & Pages section. 1
A low-risk response
- Audit new workflow definitions. Review all Cloudflare Workflows registered on or after September 10, 2026. Identify which pipelines support mission-critical data processing, financial reconciliation, or compliance audits. 1
- Apply asymmetric retention in invocation handlers. Update calls to
env.MY_WORKFLOW.create()to pass explicitretentionoptions. SeterrorRetention: "30 days"on workflows where incident root-cause analysis requires full history. 13 - Shorten success retention for high-frequency jobs. For jobs that run every few minutes or process large payloads, lower
successRetentionto 24 or 48 hours to minimize billable storage. 14 - Export critical state to object storage. If long-term compliance mandates storing workflow records beyond 30 days, stream step results directly into Cloudflare R2 or an external data warehouse rather than relying on workflow instance state. 2
- Verify dashboard settings. Check the Cloudflare dashboard to ensure retention rules match your team's operational escalation SLAs. 1
Verdict: tune retention now, migrate nowhere yet
Cloudflare's retention change is an operational configuration signal, requiring code and dashboard adjustments rather than any infrastructure migration. Teams already running Workflows benefit from grandfathered 30-day retention on existing pipelines, while high-volume deployments will see lower storage charges on new pipelines. 14
Full-stack developers and DevOps engineers should update their workflow authoring templates to declare explicit retention options. Adopting asymmetric retention gives the best outcome: lower monthly storage bills for normal runs and a full 30-day investigation window for unexpected errors. Keep your workloads on Cloudflare Workers and adjust the retention flags.
References
- 1Default instance retention for new Workflows on Workers Paid is seven days
developers.cloudflare.com
- 2Workflows limits
developers.cloudflare.com
- 3Workers API
developers.cloudflare.com
- 4Workflows pricing
developers.cloudflare.com
This story was produced automatically by a channel. One sentence is all it takes for Neodrop to keep producing for you.