“Three failures your n8n error workflow cannot catch (with real examples)”

Three failures your n8n error workflow cannot catch

If you run n8n for clients, you probably have an error workflow set up. Good — it’s the right first layer. But after spending time analysing real production execution data, here are three failure classes it misses entirely.


  1. The workflow that succeeds by doing nothing

A node with continueOnFail = true can fail silently. The execution finishes green. The error workflow never fires. n8n’s status says “success”.

The only way to catch this is to walk data.resultData.runData per node — not just check the top-level execution status. If you’re only watching for status: error, you’re missing this whole class.

This is the most expensive one in production because nobody finds out until the client notices.


  1. Your error workflow itself failing

If the error workflow has a bug — expired Slack webhook, renamed channel, rate limit — it fails trying to report a failure. Nothing reports that.

There’s no global error handler in n8n. So your alerting can go dark for days and you won’t know.

The tell: check your error workflow’s own execution history. If you see failures there, you’ve had a blind spot.


  1. “Unable to sign without access token” — the credential failure with no HTTP code

Most credential monitoring matches on HTTP 401. But Google OAuth can fail at the signing step, before the request is ever sent. The error has no HTTP code — just the message “Unable to sign without access token”.

If your classifier only checks HTTP status, it misses this. It looks like an unknown error instead of a dead credential.


None of these are edge cases — they come up regularly in real production data.

For anyone building their own monitoring: the classifier needs to flatten message + description + cause.message + httpCode into one signature before matching. HTTP code alone is not enough.

I wrote a longer technical breakdown here if useful: Why n8n says a workflow succeeded when it didn't — Okum and Already have an n8n error workflow? Here's what you're still missing — Okum

— Ashim, building Okum (the accountability layer for n8n agencies)

1 „Gefällt mir“

Two of these I hit myself when I built an Error Trigger → Gmail alert setup on self-hosted n8n and then tried to break it on purpose.

  1. The green run that did nothing. I had “Continue on error” on an AI classification step in a lead sorter. The step failed, the execution finished green, and the lead carried on with an empty category. What fixed it for me was on the workflow side, not the monitoring side: use the node’s error output instead of a plain continue, and on that branch flag the lead, so the failure is visible in the data itself and not only in the execution list.

  2. The alert workflow failing. Same here: when the alert workflow itself failed, nothing told me. My workaround was an error output inside the alert workflow that logs the failed alert to a Google Sheet, so a dead alert still leaves a trace through a second channel.

One more that surprised me: “Retry on fail” with a wrong API key doesn’t hide the error, but it delays the alert by the whole retry time. For an auth error the retries were just waiting.

I haven’t seen the “Unable to sign without access token” case myself. What I did see with Google OAuth was an app left in Testing mode: the login expired after 7 days and the workflow simply stopped.

Thanks for actually testing this on purpose — that’s the only way these show up. Two things in your reply are better than what I wrote:

  1. Retry on fail with an auth error. You’re right, and it’s easy to miss. Retries make sense for 429s and 5xx, but on a 401 they just delay the alert by the whole retry window. Worth enabling retry only on nodes that call flaky services, and keeping the wait short.

  2. Logging the failed alert to a Google Sheet as a second channel is a neat answer to “who watches the watcher”. A dead alert still leaves a trace.

On “Unable to sign without access token”: it appears when the OAuth credential has no usable token at the signing step, so n8n fails before the request is even sent. That’s why it has no HTTP code. Your Testing-mode finding is a very likely cause for Google credentials: the app is left in Testing, the refresh token dies after 7 days, and the workflow just stops. If a Google node keeps dying weekly, check that first.

Curious: is your setup for your own workflows, or for clients?

Checking runData is useful for swallowed node failures. There’s a case it still can’t settle: all nodes succeed, the destination acknowledges the write, and the destination count is right, but it contains the wrong record.

I ran four synthetic cases in n8n 2.47.7. In the wrong-record case, the workflow expected A and the destination contained B. Both counts were 1; the number of matching keys was 0. A true no-op also finished green, but expected and observed were both empty.

I put those cases in a small importable workflow. It uses a manual trigger and fictional data, with no credentials or outbound calls. For a real run, use the accepted source keys and a separate, complete destination read for the same window. If that read is missing or partial, the result stays unverified rather than counting as zero.

One more failure mode worth testing is when the workflow and its alerting both appear healthy, but the expected business outcome never happens.

For a lead-routing workflow, I’d run a synthetic lead through the same path and independently verify that the correct CRM record, owner, and next action exist within an agreed time window

That check should run outside the n8n instance being monitored. Otherwise, a stalled trigger or an unavailable instance could also prevent the health check from running.

It gives you a way to detect failures that execution status and node-level error inspection alone won’t catch

Mahdi, I tried your lead-routing example with the same import. The matching key is now the tuple (lead ID, owner ID, next action). The wrong-owner and wrong-action cases both came back as mismatches even though each read still returned one row. A complete read with no expected lead was a mismatch; an unavailable observer stayed unverified.

I added the seven fictional input cases and instructions to the same download. The workflow itself didn’t change. For the deadline you mentioned, I’d use a snapshot at the cutoff or CRM history: a later read can show the right result exists, but not that it arrived on time.

Good point on the cutoff. I’d use the CRM’s change timestamp or audit history to verify when the correct owner and next action were actually recorded, rather than when the observer first detected them. That keeps SLA compliance separate from eventual correctness

I tried that split in the same import. For one fictional lead, the next action was correct in the current view but wrong at the agreed cutoff. The import returned delivered now and mismatch at cutoff. With no complete history, the cutoff result stayed unverified.

Here’s the paired input for the same workflow. No CRM or credentials needed. On a live run I’d use the relevant field history or an actual cutoff snapshot, not a generic record updatedAt that might reflect an unrelated edit.

Exactly. A generic updatedAt can give false confidence when unrelated fields change. Field-level history or a cutoff snapshot is the stronger evidence here. Nice to see the test distinguish current correctness from SLA compliance