Where do I see whether the trigger for crash loops fired and which run it woke?

Tares shows whether the crash loops trigger fired on the trigger’s page and under Firings, and shows the run it woke in Activity and the agent’s Runs. Activity puts the firing and the run it woke in one thread. If the trigger did not fire, check that its sources and filters match the events that arrived.

How it works

A trigger evaluates as events arrive from the sources it watches. When its filters and condition match, it fires and wakes the agent subscribed to it. Tares records each step in the console, so you can follow the chain from the event to the run:

• The trigger’s page and Firings list every time the crash loops trigger fired.
• Activity shows the firing and the run it woke as one thread. You find it under Setup → Advanced setup.
• Agents → your agent → Runs has every run of the agent that the trigger woke.
• Explore shows the events on the service’s timeline, so you can see what the trigger had to evaluate.

If the agent wrote a finding, it appears on the project’s Overview under What the agents found, and on the service’s timeline in Explore. For the full picture of how sources, triggers and agents fit together, see AI SRE.

Set it up

1. Make sure Tares is running and open the console. On a local Tares, start it with the command below and open http://127.0.0.1:8787.

tares up

2. Open Triggers and pick the crash loops trigger. Its page lists its firings. You can also look under Firings for the same entries.
3. If the trigger fired, open Setup → Advanced setup and then Activity. The firing and the run it woke show up as one thread, so you can tell which run belongs to which firing.
4. To see the run itself, open Agents, pick the agent subscribed to the crash loops trigger, and open Runs. Every run of that agent is listed there.
5. If the trigger did not fire, open the trigger and compare its configuration with the events. Check these parts:
   • sources: the trigger must list the sources that carry your crash loop events.
   • filters: each filter’s field, op and value must match the events. A filter that never matches means the trigger never fires.
   • condition: the aggregate, predicate and window must be satisfied by the events that arrive.
6. Open Explore, pick the service, and confirm the events you expect are on its timeline. If the events are missing there, the problem is upstream of the trigger, in the source that should deliver them.

What you get

• A clear yes or no on whether the crash loops trigger fired, from the trigger’s page or Firings.
• The exact run the firing woke, so you can read what the agent did without searching through every run.
• One thread in Activity that ties the firing to the run.
• A way to debug a trigger that stayed quiet: because it evaluates as events arrive, comparing its sources and filters with the events on the timeline shows why nothing matched.

When something else fits

• If you want to investigate the evidence yourself instead of checking the trigger, connect an MCP client and read the correlated timeline for the service, as the AI SRE guide shows.
• If you want every firing and finding to land where your team already looks, send them to a Slack channel in addition to checking the console.

Related