Skip to content
Docs
flero.ai

Sandbox mode, troubleshooting

Common surprises when working in sandbox mode and how to resolve them.

"My HTTP node returned weird-looking data in sandbox"

Likely cause: the node is using a generic synthesised mock (priority 4). Shape matches the schema, values are random / placeholder.

Fix: record a real response by running once in production, then Save as mock, or Pin the exact data you want.

"I pinned data but sandbox ignores it"

Check:

  • The Sandbox toggle in the chrome is actually on.
  • The pin shows the green "Pinned" badge on the node's output.
  • The pinned JSON parses correctly (the badge is red if JSON is invalid).
  • You're on the correct workflow version, pins are per-version-of-the-workflow in some configurations.

If all of those check out, restart the worker (admin action), there's a known cache that occasionally needs to be flushed.

"Sandbox run took as long as a real run"

Possible causes:

  • External call latency, a Disable sandbox for this node toggle is on for one of the nodes, hitting a real (slow) API.
  • Heavy internal computation, Code nodes, Loop nodes, big Data Transform operations execute for real even in sandbox. Optimise normally.

"Approval pause is blocking my sandbox iteration"

Workflow settings → Sandbox → Auto-approve approvals. Turn on. Now Approval nodes auto-approve after a 1-second pause.

"My Stripe / Slack / Salesforce node is failing in sandbox"

Connector nodes ship mock data, but some operations don't have built-in mocks (rare / new ops). The node will fail with No mock data available for operation <X>.

Fix: record once in production (test mode if the service offers it) and Save as mock, OR pin custom data.

"Sandbox runs are showing up in monitoring totals"

By default, the monitoring dashboard excludes sandbox runs. If you're seeing them counted:

  • The dashboard's Include sandbox toggle is on. Turn it off.
  • Or you have a custom alert configured to count all runs, adjust the alert filter.

"I want a sandbox run to deliberately fail to test error paths"

Two approaches:

  • Pin error output on the upstream node, set the data to whatever shape your downstream code treats as an error.
  • Force-fail a specific node in sandbox, Advanced setting under the node's inspector: Force failure with a configurable error type.

Both leave production behaviour unaffected.

"Sandbox mode changed and my workflow stopped working in production"

Sandbox toggles are per-run by default. Make sure you haven't accidentally enabled Sandbox by default in workflow settings, the workflow status chip will show a yellow indicator if so.

"How do I share sandbox runs with a teammate?"

Sandbox executions appear in /logs like any other run, with the same shareable URL. Filter by sandbox = true to find them, then share the execution detail URL.

When to escalate

If sandbox is consistently giving wildly wrong output for a connector, and pinning doesn't fix it, file a bug. Include:

  • Connector name and operation.
  • Expected response shape (from the service's docs).
  • Actual response shape Flero produced.
  • Execution ID.

Sandbox mocks are maintained by the connector authors; out-of-date mocks happen and we fix them.

Found something out of date? This page lives in the Flero docs content set.