routes
def routes() -> List[Route]The table, first match first.
orca.web.routes
Every path orca ui answers, and what answers it: the route table orca.web.server serves (orca.web.router for how a path matches).
Each handler here only takes the request's parameters out of the Call
and hands them to the page (orca.ui) or the action (orca.api.actions)
that does the work. A page answers HTML, or its JSON to Accept: application/json or ?format=json (/api/... always JSON). The worker
API (/api/worker/...) has a table of its own, orca.api.worker, and the
stylesheet is served before any route is looked up.
| Path | |
|---|---|
/ |
pipelines, with their coverage (orca.coverage: partitions or keys done of those it has, coverage in the JSON), schedules, next runs and last runs |
/pipelines/<name> |
?tab=overview (default), runs or operate, each its own page; a pipeline's steps, partitions, schedule, next runs and runs; a keyed one's keys, done, live and failed; its fill (orca.fill): done of total, rate, ETA, today's estimated spend against its budget and why it's idle; its registration as orca config show prints it (#config, Overview); ?preview=1&partitions= (&rerun=1): what Start runs would queue, queueing nothing; ?steps=graph or table: the Steps tab shown without scripts |
/pipelines/<name>/progress |
the page's Progress card (runs going, fill, how far a sensor is behind what it follows (orca.follow), next run, coverage), as an HTML fragment the page refreshes every 5 s; ?tab= is the tab polling it, which its "Log in to ..." links come back to |
/api/pipelines/<name> |
the same as JSON, whatever is asked for |
/pipelines/<name>/calendar |
a daily or monthly pipeline's days (months) of ?month=2023-01 (2023) with their state, as JSON: done, failed, live, open, outside; the page shows it as coverage (?cal= a month or year, ?cov=month or year: a daily one's tab, year by default, both from one read of the year in orca.ui.coverage) and as the Start runs form's date picker |
/pipelines/<name>/keys/<key> |
one key of a keyed pipeline: its steps in every run it was in |
/runs |
runs, newest first, ?limit= (or ?n=; 50, at most 500) to a page; ?pipeline=, ?status= (one or more), ?trigger=, ?from=, ?to=, ?since=, ?key= (a partition: a glob or a part), ?run= (goes to that run); ?before=<id>: the page older than that run, ?after=<id>: the page newer than it (keyset, so pages stay put as runs arrive); the JSON's next, prev (and next_url, prev_url) are the cursors either side. A pipeline's Runs tab pages and filters the same, 20 to a page (runs_page in its JSON). Each status's count under the other filters is the JSON's counts |
/runs/table |
a runs list's page as the fragment it refreshes from: ?view=list (/runs), pipeline (a pipeline's Runs tab) or home (recent runs), with the list's filters, limit and cursor; X-Poll-Every says when to ask again (5 s while a run on it is queued or running, 30 s on the newest page otherwise), and an older page with nothing going on answers 286 |
/home/pipelines |
the home page's table of pipelines, as the fragment it refreshes from (5 s while a run is queued or running, else 30 s) |
/runs/<id> |
a run's cards (Status, Partitions & trigger, Code, Resources, Settings), its steps (?steps=table (default), timeline or graph: the Steps tab shown without scripts), its machines, and the snapshots each step read and wrote |
/runs/<id>/attempts/<attempt> |
an attempt's log, with its error; its facts refresh from /runs/<id>/attempts/<attempt>/facts while it runs |
/code |
?u= a link to a file at a commit on GitHub: its lines and each coloured as HTML (orca.highlight), as JSON, for the code shown on hover; the UI fetches the file, so a browser needn't reach GitHub. With &at= a line, only that def's signature, coloured, and its docstring's first paragraph, for a step's card (orca.ui.step_cards) |
/runs/<id>/timeline.json |
the run's Dista timeline, which its page draws (orca.ui.viz) and Perfetto opens |
/api/gantt |
the pipelines in dataset order with their runs over ?window=1h, 6h, 24h (default) or 7d, and the runs each started: the home page's Timeline tab (orca.ui.viz) |
/runs/<id>/driver.log |
the run's driver output, as the scheduler kept it |
/runs/<id>/metrics |
the run's machines: charts and peaks from its samples (orca.metrics) |
/ops |
redirects to /ops/alerts; "Ops" in the nav has Alerts, Control, Pools and Watchdog under it |
/ops/alerts |
Alerts (orca.ui.alerts): every alert sent (orca.alerts), newest first, 50 to a page (?limit=, at most 500; ?before=<id>, ?after=<id>), ?kind= and ?state=open or resolved; its kind, title and message, links, whether ntfy took it and whether it is still open; the newest page refreshes every 30 s from /ops/alerts/live; /api/alerts is the same as JSON |
/ops/control |
Control, the one machine running the scheduler, this UI, Postgres and local and control-placed runs (so not a pool on Pools): its own health: CPU, memory, the disk Postgres is on and network over ?range=24h, 7d or 14d, and each container's memory (orca.system), and the watchdog's banner, refreshed every 30 s from /ops/control/live; /api/system is the same as JSON; /system redirects here (301) |
/metrics |
Prometheus gauges: the control instance's CPU, memory, disk, network and containers' memory, runs running and queued, hosts, and errors in the last hour; open like the other pages (reads need no login) |
/ops/pools |
Pools, every machine runs are placed on (orca.ui.pools, orca.views.pools): a row per EC2 instance orca launched (live, and gone for an hour after it ended), per live Kubernetes run's pods (on the Kubernetes runner) and per worker token's machine (orca.workers), with its pool, kind, state, image, CPUs, memory, runs, last heartbeat, CPU and worker; ?pool= (ec2, kubernetes, worker:<pool>) and ?kind= (ec2, k8s, self-hosted) filter it; with EC2 in view, Fleet's cards (orca.fleet) and its charts over ?range=24h or 7d; refreshed every 10 s from /ops/pools/live; /api/pools the same as JSON |
/ops/pools/<kind>/<id> |
a machine: its facts, and an EC2 instance's samples over its run (the run page's Machines charts) or the runs a worker claimed |
/ops/fleet, /ops/workers |
moved to /ops/pools?kind=ec2 (Fleet's ?range= kept) and ?kind=self-hosted (301), as are /ops/fleet/hosts, /ops/workers/live, /fleet and /fleet/hosts (to /ops/pools/live for the fragments); asked for JSON, they answer Fleet's and Workers' JSON as before, as /api/fleet (the summary, hosts and charts of orca.fleet) and /api/workers (each worker token's pool, label, worker, image, last heartbeat, runs, CPUs and memory) do. A worker's own requests go to /api/worker/... with its token |
/watchdog, /api/watchdog |
the watchdog's rules with their last values, limits and snoozes, and the trips with a pipeline still paused (orca.watchdog); also the red banner on /, /ops/pools and /ops/control; the page refreshes every 30 s from /watchdog/live |
/datasets |
datasets, with their writers, readers and latest snapshots, and the graph of pipelines and datasets |
/datasets/<name> |
a dataset's place in that graph, the steps that write and read it (linked to their code), what a rebuild of it runs, and its snapshots; ?preview=1&partitions=: a rebuild's runs by level, queueing nothing |
/datasets/<name>?before=<cursor> |
its snapshots 50 to a page, newest first; ?after= for the newer page, the cursor <attempt>.<snapshot> (snapshots_page in the JSON) |
/datasets/<name>/snapshots/<id> |
a snapshot's lineage (orca.lineage): its diagram, Upstream (back in the JSON: the snapshot, if a run wrote it, and the snapshots it was written from, with the run and step that wrote each and what that step read) and Downstream (forward: the snapshots written from it) |
/rebuilds/<id> |
a rebuild's pipelines by level, with their runs (orca.sensors) |
The actions, each a POST, are orca.api.actions'; /login, /logout,
/api/login and /api/logout are orca.web.auth's.