Extracting n8n Workflow Node Execution Times and Displaying in Grafana

The Challenge

I wanted to extract accurate execution times for all nodes in all my n8n workflows, including cases where nodes run multiple times (either due to loops or because they’re sub-nodes like “Postgres Chat Memory”).

Initial Attempts

1. I started with a simple approach that matched nodes to execution times by position, but this failed for nodes that ran multiple times.
2. I then tried to identify multi-run nodes by looking for patterns in the JSON structure, focusing on:
   • Nodes with “inputOverride” in their execution data
   • Run index information (e.g., { "node": "88", "runIndex": 0 })
3. I created a code node to map node references to actual node names, but couldn’t reliably determine which node references corresponded to which nodes.

The Breakthrough

The key insight came when I analyzed the raw JSON structure and noticed:

1. In the JSON at index 2, there’s a “runData” field that points to the main node mapping.
2. Each node in this mapping has a reference number that points to an array.
3. The length of this array indicates how many times the node ran.

For example:

Copy"Postgres Chat Memory": "14"  // Points to index 14
parsedData[14] = ["23", "24"] // Array with 2 items = node ran twice

The Final Solution in Three Steps

Step 1: Node Run Analyzer

• Locate the node mapping using the “runData” reference
• For each node, check its referenced data
• If that data is an array, the length tells us how many times the node ran
• Identify nodes that ran multiple times

Step 2: Expanded Node List Builder

• Create an expanded list with one entry per node execution
• For nodes that ran multiple times, include multiple entries with different run indices
• Verify that the total matches the number of execution time objects

Step 3: Execution Time Matcher

• Match each node in the expanded list with execution time objects in order
• Output the final results with node names, execution times, and run indices
• Include a summary of multi-run nodes

The Result

So above, that’s a Grafana panel showing the min, max, and mean of each node in a selected workflow.

My next step is to show that data based on the time-range selectors in the Dashboard and then, display a chart of node performance over time 🙂