The route

Monitor: every run, in one place

Every time a pipeline runs — from the editor, from a schedule, from the API — Odara records it. Monitor is the page that lists those runs. Until now you've watched runs in the Console while you were looking; from here on, pipelines will run when you're not, and Monitor is where you find out what happened.

Goal

Find a failed run from earlier in the course, read why it failed, and re-run a pipeline from Monitor without opening it.

You'll need

  • Some history: the pipelines you built in Worlds 1–5. At least one failure helps — level 5.4's w5-4-flaky with Stop loop on first failure is perfect.
  • About 15 minutes.

Steps

1. Open Monitor

In the top bar, open the Editor menu and choose Monitor. The title in the top bar changes to Monitor.

Monitor: one row per run, newest first

Column What it tells you
Status Completed, Failed, Running, Stopped, Queued
Type Pipeline or Maestro
Name The pipeline — and, for a failed run, the start of the error under it
Started How long ago
Duration Wall-clock time of the run
Rows Rows the run wrote: its targets, plus everything written inside loops
Actions ▶ runs the pipeline again

The page refreshes itself every 5 seconds — the footer says so — so a run that starts anywhere shows up here without a reload.

2. Filter to what matters

Set the status filter to Failed and type part of a name in the search box — w5-4-flaky. Search matches any part of the name, so w5 would also catch every other name containing it. The footer shows how many runs match.

Only the failed runs, with the error under each name

The start of the error is right there in the list — Iterate node failed at row 8: Conn… already tells you which node gave up. The list cuts it to one line — hover it, or open the run, for the whole message, down to invalid Zip archive.

3. Open a run

Click a row (anywhere except ▶). The detail page shows the status, start time, duration and the execution ID; for a failed run, the full error; and the log of that run, node by node, ending with the outcome — Pipeline completed successfully (288 rows written …) or Pipeline failed: ….

The detail of one run: the error, then its log

Runs started from the editor log each node as it starts and finishes. Runs started by a schedule or the API log each node once the run is over — and when one of those fails, the log is just the failure line, which is why the full error sits at the top of the page.

4. Re-run from the list

Go back and press ▶ on a pipeline that succeeded — w1-4-filter-to-csv, for example. A new row appears at the top within a second or two. The old row stays: Monitor is history, and history doesn't get rewritten.

Expected result

  • You found w5-4-flaky among the failed runs, with the start of its error under the name.
  • ▶ produced a new Completed row at the top of the list.

Common pitfalls

Reading Rows as rows read. Rows counts what the run wrote. A pipeline that ends in a Console, or whose filter matched nothing, shows 0 — that's the number to watch for a load that silently wrote nothing (level 7.3 turns it into an alert). A loop that stopped part-way shows what its finished rounds wrote.

Re-running a pipeline that appends. ▶ is a real run: a pipeline with Append on (5.2) adds its rows again.

What you built

The habit that separates a pipeline you run from a pipeline you operate: after anything runs, look at Monitor. The next level makes things run without you.

Check yourself

One question, drawn from a larger bank. Get it right to pass.

  1. What does the Rows column count for the 5.2 loop pipeline?