prop/lib/microwaveprop/propagation/pipeline_status.ex
Graham McIntire b984794571
Hoist commercial link query + tie pipeline chip to ScoresFile
PropagationGridWorker.merge_commercial_link_data/2 called
Commercial.link_degradation_at twice per grid cell (once to count,
once to merge), and each call re-ran enabled_links plus two sample
queries per in-range link. That was ~500k SQL queries per forecast
hour — fine for the DB but catastrophic for log volume in dev
iex sessions trying to watch a chain step run.

Split into two new functions:
  * build_link_lookup/2 does all the DB work up front — one
    enabled_links query + one link_degradation per enabled link
    (~10 queries total).
  * link_degradation_from_lookup/3 is pure: takes a (lat, lon)
    and the precomputed lookup, returns the aggregate or nil.

The worker now calls build_link_lookup once per forecast hour and
the per-cell path is haversine-only. Net: 500k queries → ~10.

PipelineStatus freshness detection moves off oban_jobs.completed_at
onto ScoresFile.latest_valid_time(). The on-disk files ARE the data
the map renders from, so the chip's "Up to date · Nm ago" and the
2h stale threshold now track actual data presence. Running-state
detection still comes from oban_jobs since the chain step is still
an Oban row. Tests updated to seed a ScoresFile instead of a
completed Oban row for the idle/stale cases.
2026-04-14 15:01:12 -05:00

135 lines
4.1 KiB
Elixir

defmodule Microwaveprop.Propagation.PipelineStatus do
@moduledoc """
Aggregated status for the propagation update pipeline.
Two inputs drive the chip:
* **Running detection** — queries `oban_jobs` for executing
PropagationGridWorker / AsosAdjustmentWorker rows so the chip
knows when a chain step is in flight.
* **Freshness detection** — reads the newest valid_time from
`Microwaveprop.Propagation.ScoresFile` on disk. The on-disk
files ARE the propagation data now, so the chip's "Up to date ·
Nm ago" and the 2h stale threshold track actual data presence
instead of Oban's `completed_at` timestamp.
"""
import Ecto.Query
alias Microwaveprop.Propagation.ScoresFile
alias Microwaveprop.Repo
@grid_worker "Microwaveprop.Workers.PropagationGridWorker"
@asos_worker "Microwaveprop.Workers.AsosAdjustmentWorker"
@workers [@grid_worker, @asos_worker]
# Scores older than this flip the chip to :stale. Matches the
# FreshnessMonitor threshold — anything past 2h is already a gap
# the pipeline should have caught.
@stale_after_minutes 120
@type state :: :running | :idle | :stale | :unknown
@type worker_tag :: :grid | :asos
@type running_detail :: %{worker: worker_tag(), label: String.t()}
@type t :: %{
state: state(),
label: String.t(),
details: [running_detail()],
last_update_at: DateTime.t() | nil
}
@doc """
Format a forecast-progress payload (as broadcast by
`PropagationGridWorker`) into a short detail line for the running
chip ("now", "+3h", …).
Returns `nil` when the payload has no `forecast_hour` so callers can
fall back to the base detail (e.g. "HRRR run") from `current/0`.
"""
@spec running_detail(map() | nil) :: String.t() | nil
def running_detail(%{forecast_hour: 0}), do: "now"
def running_detail(%{forecast_hour: fh}) when is_integer(fh) and fh > 0, do: "+#{fh}h"
def running_detail(_), do: nil
@spec current() :: t()
def current do
case running_workers() do
[] ->
build_idle_or_stale(latest_data_at())
details ->
%{
state: :running,
label: "Updating propagation",
details: details,
last_update_at: latest_data_at()
}
end
end
# Returns every pipeline worker currently in executing state, in a
# stable grid-before-asos order so the chip sub-lines don't flicker
# as jobs come and go.
defp running_workers do
from(j in "oban_jobs",
where: j.state == "executing" and j.worker in ^@workers,
distinct: j.worker,
select: j.worker
)
|> Repo.all()
|> Enum.sort_by(&worker_sort_key/1)
|> Enum.map(&to_running_detail/1)
end
defp to_running_detail(@grid_worker), do: %{worker: :grid, label: "HRRR run"}
defp to_running_detail(@asos_worker), do: %{worker: :asos, label: "ASOS nudge"}
defp worker_sort_key(@grid_worker), do: 0
defp worker_sort_key(@asos_worker), do: 1
defp worker_sort_key(_), do: 99
# Freshness is derived from the newest ScoresFile on disk. That file
# is the thing the map actually renders from, so its valid_time is
# the most honest answer to "when was the data last updated?".
defp latest_data_at do
ScoresFile.latest_valid_time()
end
defp build_idle_or_stale(nil) do
%{state: :unknown, label: "Propagation status unknown", details: [], last_update_at: nil}
end
defp build_idle_or_stale(%DateTime{} = last) do
age_minutes = max(DateTime.diff(DateTime.utc_now(), last, :minute), 0)
if age_minutes > @stale_after_minutes do
%{
state: :stale,
label: "Propagation data stale · last update #{format_age(age_minutes)} ago",
details: [],
last_update_at: last
}
else
%{
state: :idle,
label: "Up to date · #{format_age(age_minutes)} ago",
details: [],
last_update_at: last
}
end
end
defp format_age(0), do: "just now"
defp format_age(1), do: "1m"
defp format_age(n) when n < 60, do: "#{n}m"
defp format_age(n) do
hours = div(n, 60)
mins = rem(n, 60)
if mins == 0, do: "#{hours}h", else: "#{hours}h #{mins}m"
end
end