prop/lib/microwaveprop/weather/mrms_cache.ex
Graham McIntire f1846c0a53
perf: reduce per-pod RSS and HRRR chain wall time
Telemetry showed the application-master process holding ~830 MiB of
terms from warm_grid_cache_from_latest_profile — the data lives in
the app master's heap and never GCs because the process is idle.
Running it in a Task.start lets the terms die with the task.

Mark GridCache, MrmsCache, NexradCache, and ScoreCache ETS tables
:compressed. The scored-band-map and HRRR grid data are map-heavy;
compression trims hundreds of MiB at a few percent CPU cost.

Memoise HRRR .idx responses in Microwaveprop.Cache. Published idx
files are immutable for a model run, but the hourly chain re-fetches
the same URL dozens of times across forecast hours. Cuts ~10s per
repeat out of hrrr_fetch_idx.

Force a garbage collect at the end of HrrrFetchWorker.perform to
reclaim the refc binary heap held from GRIB2 ranges before the Oban
producer hands the process its next job.
2026-04-19 14:56:48 -05:00

80 lines
2.3 KiB
Elixir

defmodule Microwaveprop.Weather.MrmsCache do
@moduledoc """
Node-local ETS cache of the latest MRMS PrecipRate grid, regridded onto
the 0.125° propagation grid. Mirrors the pattern used by
`Microwaveprop.Propagation.ScoreCache` and
`Microwaveprop.Weather.GridCache` — a single GenServer owns the table,
callers read directly from ETS, and `broadcast_put/2` fans updates out
to peer nodes via PubSub.
Only one node runs `Microwaveprop.Workers.MrmsFetchWorker` per cron
tick (Oban Pro Smart engine picks a leader), so the other nodes get
the grid via the `"mrms:cache"` topic and stay in sync without each
one paying the 1 MB fetch + wgrib2 regrid cost.
"""
use GenServer
alias Phoenix.PubSub
@table :mrms_cache
@topic "mrms:cache"
@pubsub Microwaveprop.PubSub
@type rain_grid :: %{{float(), float()} => float()}
@spec start_link(keyword()) :: GenServer.on_start()
def start_link(opts), do: GenServer.start_link(__MODULE__, opts, name: __MODULE__)
@doc "Return the cached `{valid_time, rain_grid}` or `:miss`."
@spec fetch() :: {:ok, DateTime.t(), rain_grid()} | :miss
def fetch do
case :ets.lookup(@table, :current) do
[{_, valid_time, grid}] -> {:ok, valid_time, grid}
[] -> :miss
end
end
@doc "Current MRMS valid_time if anything is cached."
@spec valid_time() :: DateTime.t() | nil
def valid_time do
case fetch() do
{:ok, vt, _} -> vt
:miss -> nil
end
end
@spec put(DateTime.t(), rain_grid()) :: :ok
def put(valid_time, grid) do
:ets.insert(@table, {:current, valid_time, grid})
:ok
end
@doc "Insert locally and broadcast to peer nodes."
@spec broadcast_put(DateTime.t(), rain_grid()) :: :ok
def broadcast_put(valid_time, grid) do
put(valid_time, grid)
PubSub.broadcast(@pubsub, @topic, {:mrms_cache_refresh, valid_time, grid})
:ok
end
@spec clear() :: :ok
def clear do
:ets.delete_all_objects(@table)
:ok
end
@impl true
def init(_opts) do
:ets.new(@table, [:set, :named_table, :public, :compressed, read_concurrency: true])
PubSub.subscribe(@pubsub, @topic)
{:ok, %{}}
end
@impl true
def handle_info({:mrms_cache_refresh, valid_time, grid}, state) do
put(valid_time, grid)
{:noreply, state}
end
def handle_info(_msg, state), do: {:noreply, state}
end