Debug with Niro
When something fails, you want facts about the code around the failure: where an error comes from and who catches it, who writes a wrong value, which service answers a failing call, what a change reached, which setting a call depends on, and what a slow method does. Niro answers each of these from the code. The graph answers are the same every time on the same graph; a text search (code_grep) can stop early on a busy machine, and then says PARTIAL and how to carry on. An exception is an error a method throws, which a caller further up may catch.
Pick the playbook by symptom
| What you see | First tools (with the short answer form) |
Playbook |
|---|---|---|
| An error message or a stack trace | code_grep for the message (answer=locations), get_source_code, get_exception_flow both ways |
Playbook: debug an error message or a stack trace |
| A wrong value in a table or a response | get_data_resource_usage, code_grep for the setter (answer=names), get_impact_analysis (answer=names) |
Playbook: debug a wrong value |
| A call to another service fails | find_outbound_http_calls, find_handler, get_service_topology, niro route-rule unconnected |
Playbook: debug a failing call between services |
| It broke after a change | get_blast_radius with method_fqns (answer=count, then names) |
Playbook: "it broke after my change" |
| It works here and not there | niro env list, code_grep over config files, niro route-rule unconnected |
Playbook: debug config and environment |
| A path is slow | get_call_chain direction=callees (answer=count, then full), get_data_resource_usage |
Playbook: debug a slow path |
| A message was sent and nothing happened | get_event_producers_and_handlers, find_orphaned_events |
Playbook: data stores and events |
Every playbook shows the real answers, says why they are complete or where they are not, and compares them with grep and file reads.
How
niro ask get_exception_flow method_fqn=PaymentService.charge
niro ask get_exception_flow method_fqn=PaymentService.charge direction=upstream exception_type=TimeoutException
- Downstream (the default) lists the exceptions that can leave the method, and where each is thrown and caught, 3 levels deep by default and 5 at most.
- Upstream lists which callers must handle a given exception. It needs
exception_type. Without it, the answer is "Error: direction=upstream needs 'exception_type'".
Two other tools help while debugging:
get_call_chainwithpurpose=DEBUGadds the source code of the method you ask about.find_handlertells you which code answers a call to another service.
Known limits
- An exception that is caught and wrapped in another can still be listed as escaping.
- A catch is matched only by the exact exception name.
Root cause analysis
Root cause analysis from your alerts (for example from Sentry, Slack or Jira) is a hosted Niro feature. The local graph does not answer it. See Features of hosted Niro.