[🐛 Bug]: workflow enrich data in 80%

[00041275]

I have bug for workflow in my keephq which have 4 instances with redis Enrich data apply for alert in 80% cases, and if I call workflow manually its work. I added python provided for time sleep(15) but bug is steel, and I haven't error in logs How to decide this problem?

This is log with alert enrich, but alert haven't enriched: 2025-08-08T04:35:59.026410: Running step description2 2025-08-08T04:35:59.026702: Action description2 evaluated to run! Reason: no condition, hence true. 2025-08-08T04:35:59.026881: Checking throttling for action description2 2025-08-08T04:35:59.027124: Action description2 is not throttled 2025-08-08T04:35:59.029590: Running description2 StepType.STEP, current retry: 0 2025-08-08T04:36:04.032499: Enriching alert 2025-08-08T04:36:04.032931: Enriching disposable enrichments 2025-08-08T04:36:04.096142: Alert enriched 2025-08-08T04:36:04.097777: Step description2 ran successfully

[dosubot[bot]]

This is a known issue in keephq, especially in distributed setups with multiple instances and Redis. The root cause is a fragile connection between workflow step context and enrichment logic—when workflows include steps, context keys may not be populated as expected, causing enrichment to silently fail even if logs show success. This is often triggered by the presence of steps in the workflow, not by the enrichment definition itself. Manual execution works because the context is correct, but in automated/distributed runs, race conditions or context sync issues can cause enrichment to be skipped without errors in logs (details, details).

To troubleshoot: try simplifying your workflow to use only actions (no steps) and see if enrichment always applies. If so, the issue is almost certainly context handling between steps and actions. Also, check your Redis and deduplication settings (KEEP_DEDUPLICATION_DISTRIBUTION_ENABLED). If possible, test with a single instance to confirm if the problem is related to distributed coordination.

There is no confirmed fix for this core bug yet. As a workaround, ensure each action/step gets its own copy of data (especially in foreach loops), and avoid shared state between steps (foreach bug). If you need to enrich alerts reliably, keep the workflow as flat as possible or trigger enrichment from actions directly.

Recent fixes (like PR #5231) addressed other enrichment bugs, but not this context-handling issue. If you need more technical details or want to track the issue, see issue #4958 and issue #4799.

To reply, just mention @dosu.

---

^{How did I do? Good | Irrelevant | Incorrect | Verbose | Hallucination | Report 🐛 | Other}