✨ Troubleshooting Agent

Overview

The Troubleshooting Agent will help your teams find out why a Spark job failed or why a table's data is stale, and what to do about it. It will run inside your IOMETE deployment and see only what the asking user can see.

What it will look at

Failed Spark jobs

• A job run's configuration, and its driver and executor logs.
• Errors, warnings and failure patterns across those logs, summarised into a likely cause.
• For runs whose driver never started, the run's error and its Kubernetes events.

Stale or missing data

• When a table was last loaded, compared with how recent its data is.
• Load volume over recent days, to spot a drop.
• A comparison of a table with the source table you name, to find where rows went missing.

Access and connections

• For administrators: why a user cannot see a catalog, table or compute cluster, from the user's roles, resource bundles and data security policies.
• Why a BI tool or JDBC client cannot connect, from the compute cluster's connection settings.

How it will work

• Read-only. It will explain what went wrong and suggest a fix; it will not change jobs, compute clusters or policies.
• It will be available in the IOMETE console and through MCP, where it will add its own tool group.
• It will reach models only through the Model Gateway.

Where it fits

The Troubleshooting Agent builds on the Model Gateway and comes after Workload Optimization, which covers jobs that ran slowly. Its first tools will arrive with MCP Server for AI agents V2.