* perf(routes): batch cross-IATA and hash route search (#216)
One search fired a /routes/cross request per IATA pair (138k on 2026-10-08). Route search now takes an optional iatas list and runs a fixed number of index-backed queries. /info advertises built-in minimum client versions.
* chore: bump version to 2.0.4
The cursor was the last row's last_seen in ms, so routes sharing that
millisecond (common within one upsert batch) never came back. GET /routes
takes an optional cursorId, backed by keyset indexes; cursor alone keeps
working as before.
GET /traces takes an optional cursorTag so tags sharing a millisecond survive
paging. Migrations run under an advisory lock and record each file in the
same implicit transaction, so a crash can't strand a fresh database.
Overview windowHours counts the hours it covers, hourly observer activity
fills unrolled hours from raw rows, and integration fixtures are UTC.
Stats endpoints read hourly rollup tables filled once per closed hour by one
advisory-locked task, instead of materialized views that re-scanned every
observation on each refresh. IATA-set tables give exact distinct counts for any
IATA set or region. Cleanup holds raw rows for unrolled hours (up to 24h).
Adds /stats/series; ingest keeps packet observation counts, observer last IATA,
channel message counts and trace tag summaries. 001_baseline.sql is rewritten,
so pre-release 2.0 databases must be recreated. See docs/historical-stats.md.
Bucketed airtime and receive errors used MAX - MIN per bucket, so one zeroed,
replayed or garbage row counted the whole running total (e.g. 47,737 s RX in a
6 h bucket). Buckets are now summed from per-row increments with dips, stale
replays and physically impossible jumps rejected. Ingest skips uptime-only
stats and a migration deletes the rows already stored.
A repeater that advertises 0/0 is hiding its position. It was stored as a literal 0/0 that every consumer had to special-case; now the stored latitude, longitude and location source are cleared, an omitted location still preserves the stored one, and the nodeUpdate event sends explicit null lat/lng so clients can clear too. Migration 043 drops rows stored as 0/0 before this change.
Storing a full candidate snapshot on every observation was the wrong call. The same node lists got copied thousands of times, about 22 GB on prod. Resolve endpoints live like path hops and drop the column.
The payload-breakdown and top-observers matviews stored a single 7-day
total per group, so the 24h/7d/30d selector did nothing: payload ignored
the window entirely (the store discarded `since`) and top-observers only
varied which rows cleared the last_heard cutoff, never the counts.
Bucket both views by hour over a 30-day horizon and sum the buckets in
range, mirroring mv_hourly_iata_stats. The stats endpoints now honour the
requested window while still reading a small precomputed table.
GetScopeStats left-joined packets, observer_scopes and nodes at once,
multiplying rows into the millions before COUNT(DISTINCT) deduped them
(~10s per call). Count each table on its own and index packets(scope_id).
Refresh last_heard on duplicate observations too, capped at hourly, so
steady traffic can't age a trace out of the filter (dedup key has no
heard_at). Drop the unused last_heard index column so upserts stay HOT,
and share one retention cutoff across the cleanup deletes.
The trace list joined every trace packet to ~50M observations to apply
the IATA filter; the parallel hash join overran docker's 64MB /dev/shm
and the filtered list errored. Keep a small trace_iatas table at ingest
(like channel_iatas) and group packets by tag instead. Filtered stats
are now per tag rather than per matching packet.
Refresh last_heard on duplicate observations too, capped at hourly, so
steady traffic can't age a channel out of the filter while its packets
stay retained (dedup key has no heard_at). Guard the seed join against
the docker /dev/shm cap like the trace one, and drop the unused
last_heard index column so the upserts stay HOT.
The IATA filter ran a correlated EXISTS over packets/observations with
ILIKE, which skipped the iata index and took ~7s live. Keep a small
channel_iatas table at ingest (like node_iatas) and filter against it.
Also honor the iatas= param so multi-site regions stop getting the
global list. Filter now ages out with packet retention rather than
matching any retained packet.
The packet list sorts by last_heard_at, so every page was a seq scan
plus sort over all packets. Cheap to maintain now that presence writes
are coalesced.
this was causing duplicates and each obersver heard a given
packet hash or it didn't, there are not more observations of
the same hash from one observer
MrAlders0n