Monitoring workflow runs and versions
Track every workflow run, debug and re-run failures, and manage draft, active and archived versions so your Vetta automations stay reliable.
Updated October 25, 2026 · 7 min read
Once a workflow is active, Vetta runs it every time its trigger fires, and keeps a record of each run and each version. You get two levels of visibility: Global Analytics, a workspace-wide overview across all workflows, and the Runs and Versions tabs of each workflow. Together they tell you whether your automations are healthy, why a run failed, and exactly which version did what.
The workflows overview
On the Workflows page, click Global Analytics to see automation health across the workspace (and View All Workflows to go back to the list):
- Run health: total runs, failed runs and the fail rate over the period
- Needs attention: any workflow with a failed run among its recent runs, with its last failure and top cause
- Failures by cause: failed runs classified by error type
- Runs & outcomes: recent runs and their results
- Token consumers: which workflows use the most AI tokens (relevant when you use the AI CRM Agent step)
Use this page as a daily health check before you dig into an individual workflow.
Workflow runs
Each run is one execution of a workflow. A run moves through these statuses:
- Not started: created but not yet picked up
- Enqueued: waiting to run
- Running: in progress
- Completed: finished
- Failed: stopped with an error
Open a workflow's Runs tab to see how it's doing: how many runs completed, failed or are in progress, and the average run time. Filter the list by status or by date and time, and open any run to see which steps it went through and which version it used.
Execution logs
Each run has a step-by-step log. Click any step in a run to inspect:
- The trigger event details
- Which steps ran, and their status
- The input each step received
- The output each step produced (available to later steps as variables)
- The error, if the step failed
In the editor, variables read null because nothing has run yet. In a run, every variable shows the real value that run used, which is how you check that a step got the data you expected.
Testing and debugging
A workflow has no test button: you check it on real runs. The one exception is the HTTP Request step. Its Test tab lists every variable the request uses: type a test value for each and click Send Request. The request goes out from your browser and the response comes back with its status and time. Nothing is saved as a run.
Once the workflow is live, Runs is where you debug:
- Filter the list to Failed and open a run. The step that broke has a red cross.
- Click it. Its output shows the error, for example a 503 from another app.
- Fix the cause, then click Re-run.
Re-run has three options:
- Retry from failed step: keeps the steps that already worked and runs the rest again (only for failed runs)
- Restart from beginning: runs every step again
- Re-run on latest version: runs it on the workflow's newest version instead of the one the run originally used
Turn on Create a new run to keep the original run untouched in the history, which is useful for auditing what failed. Leave it off and the same run is reset and re-executed in place, which keeps the runs list tidy.
Debugging a workflow. There is no test button, so you check real runs. The exception is the HTTP request, whose Test tab lists its variables, takes test values and sends the request from your browser, showing status and time without saving a run. Once live, filter Runs to Failed, open the run, click the step with the red cross to read its output error, fix the cause and click Re-run. Retry from failed step keeps what already worked.
Transcript
A workflow has no test button: you check it on real runs. The one exception is the HTTP request. Its Test tab lists every variable the request uses. Type a test value for each, and click Send Request. The request goes out from your browser, and the response comes back with its status and time. Nothing is saved as a run. Once the workflow is live, Runs is where you debug. Filter to Failed, and open a run: the step that broke has a red cross. Click it, and its output shows the error, here a 503 from the billing tool. When the cause is fixed, click Re-run. Retry from failed step keeps the steps that already worked and runs the rest again. Back on All, the new run has completed. Every morning, Alex asks Claude how last night's runs went. Claude finds four failed runs, all at one step. Claude, ChatGPT or any assistant you connect to Vetta works on your real workspace, so ask daily, then give it a bigger job each week. In Vetta, open Runs, filter to Failed, then click Re-run.
A green run is not proof the data is right. An expression with a mistake comes out blank instead of failing, and a variable that points to nothing becomes empty text, while the step still shows as completed. After any change, open one run and check the actual input and output of the steps that matter.
Workflow versions
Vetta keeps every version of a workflow, so you can always see what ran when. A workflow version is in one of these statuses:
- Draft: being edited; never runs
- Active: the version that runs when the trigger fires
- Deactivated: turned off with Deactivate; kept in history
- Archived: an older version replaced by a newer one
When you change a live workflow, for example by adding a step, the live version stays untouched. Your change starts a new draft, version 2, and version 1 keeps running until you activate the new one. Then version 1 is archived, and the Versions tab keeps the whole history. A run that had already started when you activated version 2 finishes on version 1; only new runs use version 2.
With version history you can:
- Review previous versions and see how the automation evolved
- See which version each run used when you debug
- Open an earlier version and choose Use as draft (confirm with Override Draft) to roll back a bad change
- Discard Draft to abandon unpublished edits and keep the active version as it is
Making a workflow live and keeping track of it. Activate appears once the workflow has a trigger and at least one step; confirming puts version one live. Editing then starts a new draft, version two, while version one keeps running until you activate the new one, which archives the old. Versions keeps the history. Runs shows completed, failed, in progress and average run time, and each run shows its steps. Re-run can retry from the failed step, restart, or use the latest version. Deactivate stops it.
Transcript
Nothing runs until you activate it. Once a workflow has a trigger and at least one step, Activate appears. Click it and confirm, and version one goes live. Now change something, like adding a step, and the live version stays untouched. It starts a new draft, version two, and version one keeps running until you activate the new one. Then version one is archived, and Versions keeps the whole history, so you can always see what ran when. Open Runs to check how it's doing: completed, failed, in progress, and the average run time. Filter the list, and open any run to see which steps it went through. Re-run can retry from the failed step, restart from the beginning, or run it again on the latest version. And Deactivate stops it whenever you need to. Each release, Alex asks Claude to activate and keep a baseline. Claude reads 112 runs, then activates version three. Claude, ChatGPT or any assistant you connect to Vetta works on your real workspace, so make it routine, then give it a bigger job each week. In Vetta, check Versions, then open Runs next week.
Notifications
When a workflow reaches a Form step, the person who has to fill it in gets a Form Input Needed notification saying the form is waiting for their input. See Notifications in Vetta.
For failures, make Global Analytics part of your routine: the Needs attention panel lists every workflow as soon as one of its recent runs fails.
Common questions
Why did my run fail, and how do I find the step that broke?
Open the workflow's Runs tab, filter to Failed and open a run. The step that broke has a red cross; click it and its output shows the error, for example a 503 from another app. See Testing and debugging.
Can I re-run a failed run without repeating the steps that worked?
Yes. Fix the cause, click Re-run and choose Retry from failed step. Turn on Create a new run if you want the original kept in the history for auditing. See Testing and debugging.
Why does my run show as completed but the field it wrote is blank?
A green run is not proof the data is right: an expression with a mistake comes out blank instead of failing, and a variable that points to nothing becomes empty text. Open the run and check the actual input and output of the step. See Testing and debugging.
Why do all the variables show null in the editor?
Nothing has run yet. Open a run from Runs and every variable shows the real value that run used. See Execution logs.
Can I test a workflow without running it for real?
No. A workflow has no test button, so you check it on real runs. The one exception is the HTTP Request step, whose Test tab sends the request from your browser without saving a run. See Testing and debugging.
What happens to the live version when I publish a new one?
The live version keeps running until you activate the draft. Then it becomes Archived, the new version takes over, and the Versions tab keeps the whole history, with each run showing which version it used. See Workflow versions.
How do I roll back to an earlier version?
Open the earlier version in the Versions tab and choose Use as draft, then confirm Override Draft. Activate that draft to make it live again. See Workflow versions.
How do I throw away edits I don't want?
Click Discard Draft. The unpublished edits are dropped and the active version stays exactly as it is. See Workflow versions.
Does a run that is already in progress finish on the old version when I activate a new one?
Yes. A run that already started finishes on the version it started with, and new runs use the new version. See Workflow versions.
What failure threshold puts a workflow in the Needs attention panel?
A single failure is enough: a workflow shows there as soon as one of its recent runs failed. See The workflows overview.
Key outcome
You can now see at a glance whether your automations are healthy, find the exact step and input behind any failure, re-run it once the cause is fixed, and roll back to an earlier version if a change went wrong, all while the active version keeps running untouched.
Still stuck?
We reply the same business day, 9:00 to 17:00 CET, Monday to Friday.