Diagnose workflow failures by symptom

26R1 reference · Documentation-based tutorial draft · Not executed in IFS

Capture the smallest useful evidence set

Record process key, active version, trigger type, timing, timestamp, execution user, input identity, expected outcome, and actual outcome. Include the first failing task or async job ID. Remove credentials and sensitive payload values before sharing a report.

Symptom guide

Symptom	First checks	Next evidence
Nothing runs	Deployed version; enabled/published reaction; exact operation match	Request trace and same-user Watch
Observation list is empty	Same user starts Watch and triggers; watch window still open	Repeat one controlled request
Wrong branch in Inspect	Missing inputs, quoted booleans/numbers, enum type differences	Variables immediately before the gateway
Resource not found	Complete key/filter; row-level visibility; actual record existence	Read under the same role
ETag is incorrect	Read/update identity and ETag variable pairing	Values from the last read and current record
Script changes a value but it is not saved	Enrichment registration, writable attribute, entity result	Final payload and persisted record
Deployment succeeds but command omits it	Template state and command eligibility	Explicit command validation
Work persisted after cancellation/failure	Form, REST, autonomous, or explicit-commit boundary	Database and target-system state
Async request is POSTED but no result	Execution status, queue age, retry history, version review	Asynchronous Workflow Executions
Works as administrator only	Effective business-role grants and metadata access	Test with intended role
Inspect fails with projection-access error	Direct API grants needed for inspection	Compare normal deployed execution
Previously working access is invalidated	Changed activities after deployment	Permission-set review state
REST call fails parsing	Content type, body shape, JSONPath, empty response	Sanitized response contract
Workflow runs twice	Multiple reactions, events, or API cascades	Ordered request trace
User-form instances remain open	Abandoned form or form reached from async cascade	Open tasks and runtime data

Investigate a missing-reference failure

For the first validation tutorial, check the variable name at the trigger, its type and value before the script, the computed referenceMissing, the gateway expression, and the actual persisted result.

If the real attribute name differs from the teaching alias, fix the explicit mapping. Do not change the gateway until the input contract is correct.

Investigate an API failure without hiding it

Temporarily enabling Log Errors changes exception handling. It must be paired with a branch that checks the recorded error. Otherwise the workflow may finish successfully with missing output.

Inspect whether the issue is missing data, row security, an unsupported API contract, or concurrency. These require different fixes; none is resolved reliably by broadening permissions or turning every error into success.

Investigate asynchronous failure

Read the complete error history, not only the most recent error. Check whether the job was posted under another version and whether a manual rerun already occurred.

Before rerunning, inspect side effects from earlier attempts. A successful external write followed by a timeout is not the same as a request that never reached the target.

Escalate with a reproducible case

Supply a minimal redacted BPMN/export, trigger configuration, release, role, input case, trace/error history, and business-state evidence. State whether the problem reproduces in Inspect, the real trigger, or both.

Product incidents belong in your organization’s IFS support process; this page covers this site’s own troubleshooting guidance only.