Find what went wrong (Alerts) and what you changed (Changes)¶
Goal: see what failed or needs a look on the context (worker failures, degraded feeds, limit breaches, halts, the api going down, your failed actions), and review every write you sent, on every context.
Prerequisites
- Signed in (any user).
- For Alerts: the context in question is active, and the app has been open while things happened — alerts are collected in this browser from the moment the app opens; a reload starts empty.
Steps: alerts¶
- Notice the badge:
⚠ Nin the header counts unread alerts (red when one is an error). Click it, or open Infra > Alerts (route/infra/alerts). Opening the view marks them read. - Read the table: Last, Severity (
error,warning,info), Context, Source (worker,risk,halt,context,app), Subject, Text, Count, First. Newest first; a repeat counts up in one row and moves to the top. - Filter, e.g.
severity:error,source:worker, a worker id. - Follow up by source:
worker— a worker failed, reporteddegraded/disconnected, or had an input gap: Infra > Workers, filter by the subject (the worker id), Detail → Last failure and Lifecycle since connecting; for feeds Infra > Connections → Refresh status.risk— a limit breach of a scenario: Trading > Risk with that scenario.halt— trading halted or resumed: Trading > Scenario.context— the api went down (and an info alert when it answers again): the context menu's dot.app— an action of yours failed: the text has the reason; Infra > Changes has the entry.
- Dismiss the selected alert, or Clear all.
Steps: changes¶
- Open Infra > Changes (route
/infra/changes). - Read the table: Time, Context, What (the action, e.g. Start worker
binance-orderbook), Outcome (done,failed,declined), Detail (the reason of a failure), Saved (yes,saving…, ornot saved: <reason>). Newest first, every context together. - Filter, e.g.
outcome:failed,context:demo. - Read again to reload (e.g. after working in another browser).
An action of several calls (a bulk action, a feed connect, the kill switch) is one entry;
each operation a script or the assistant runs is one entry. A No in the
protected-context dialog is declined.
What you should see¶
- Alerts as they happen, while the app is open.
- Changes from every browser you used: the log is kept per user on the server, in each
context's state (
cfg.changelog.<USER>, the newest 500 per context).
Troubleshooting¶
| Symptom | Cause and fix |
|---|---|
| Alerts empty after a reload | By design: alerts are kept in memory only. Look at the workers' lifecycle in Infra > Workers → Detail. |
| No breach alerts | They come from sc.<S>.risk.limit.breach of the scenarios in the registry, raised by the P&L worker; no P&L worker, no breaches. |
Changes: <context>: not read (…) |
That context did not answer within 5 s; its entries are missing until it does (Read again). |
| Changes: Saved says not saved: … | The server could not store it (the reason follows); it stays for this session only. saving… is shown while it is being stored. |
| Changes: Reading the change log failed: … | The app server did not answer GET /changes. |
More: Views > Alerts, Views > Changes.