What should I do when an AI SRE run on a disk filling up ends as exhausted?

Tares ends a run as exhausted when the agent uses up its model rounds before it can conclude, and the fix is to raise max rounds for that agent. Set it under Advanced on the agent form or as max_rounds in catalog YAML, anywhere from 1 to 24. The run keeps the last text as a partial note, so you still see where the investigation got to.

How it works

Every Tares agent runs on a fixed budget of model rounds, set by max_rounds. The agent for your disk filling up trigger follows the same rules as any other agent.

• Range and default: max_rounds takes a value from 1 to 24. Unset, it is 6, or 12 once the agent uses external MCP servers. A run that reads a diff or a file and writes back needs more than the read-and-conclude budget.
• Last call: when the budget runs out mid-investigation, the model gets one final call with tools disabled. It must conclude from what it has already gathered.
• Exhausted: if the model still cannot conclude, the run ends as exhausted. Tares keeps the last text as a partial note and tells you to raise max rounds.
• Cap on record: every run records the cap it was held to and shows it as rounds/max, so you can see how close it came.
• Token limit: each call is also limited to 2,048 tokens. The round count is the setting you change.

Set it up

1. Open the agent’s page and go to Runs. Find the run that ended exhausted and read rounds/max to see the cap it was held to. Read the partial note to see which part of the investigation used the rounds.
2. Choose a new value. Raise it step by step, for example from 6 to 12. The ceiling is 24.
3. In the console, open the project’s Advanced setup, edit the agent, and set max rounds under Advanced.
4. Or, if you manage agents as a catalog, add max_rounds to the agent in catalog YAML:

agents:
  - name: incident-first-look
    project: Default             # optional; the project that made it
    trigger: incident            # the trigger that wakes it
    enabled: true                # on in its project
    max_rounds: 12
    prompt: |
      You are an SRE taking the first look when an alert fires on this service. You are handed the
      correlated timeline: the firing alert plus the signals behind it and the request logs.
      Produce a tight incident note: what is failing and since when, the most likely cause tied to
      specific evidence lines, and a suggested next action. Do not speculate beyond the evidence.

5. Let the trigger fire again, then check the agent’s Runs list. The new run shows its rounds against the new cap, and the outcome column shows how it ended.

The agent runs at most once per the trigger’s cooldown per entity, and at most 50 runs per agent per day. Wait for the cooldown to pass, or fire the trigger again after it, before you judge the new setting. For the full loop from trigger to finding, see AI SRE.

What you get

• A run that has enough rounds to read the evidence and conclude, so it ends with a finding instead of a partial note.
• A clear record of every run. The status, rounds, duration and errors sit under the agent’s Runs list, with rounds/max on each one.
• A partial note on any run that still runs out, so the work done so far is not lost and the note tells you to raise max rounds.
• One setting to change. The prompt is yours to tune per agent, and the runtime budget stays fixed apart from max rounds.

When something else fits

• If the agent keeps ending exhausted even at a high cap, tighten the prompt so it asks for less. A shorter, more focused instruction needs fewer rounds, and the partial note shows where the rounds went.
• If raising the cap does not fit, check whether the agent uses external MCP servers. That already lifts the default to 12, so an agent you set to a lower value may be capped below it.
• The cap is 24 rounds, so a task that needs more than that is better split across agents. Read how Tares agents can hand work to each other.

Related