db: squash migrations

README.md

@@ -1574,10 +1574,13 @@ jamtart/
 │   └── bin/
 │       └── tart-dash.rs        # Terminal UI dashboard
 ├── migrations/
-│   ├── 001_postgres_schema.sql             # Initial schema
-│   ├── 002_performance_indexes.sql         # Performance indexes
-│   ├── 003_frontend_analytics_indexes.sql  # Analytics indexes
-│   └── 004_frontend_search_indexes.sql     # Search indexes
+│   ├── 001_core.sql                # Extensions, raw events hypertable, nodes, event_types
+│   ├── 002_count_tables.sql        # 14 per-group count tables (all 115 event types)
+│   ├── 003_count_aggregates.sql    # _1m/_1h continuous aggregates over count tables
+│   ├── 004_union_views.sql         # all_event_stats_30s/1m/1h + all_core_stats_1m
+│   ├── 005_node_stats_services.sql # node_stats + event_services + their aggregates
+│   ├── 006_trackers.sql            # convergence, wp_tracking, DA stats & latency hists
+│   └── 007_onchain.sql             # on-chain core/service/validator stats
 ├── tests/
 │   ├── types_tests.rs          # Type system tests
 │   ├── events_tests.rs         # Event encoding tests

migrations/002_event_types.sql → migrations/001_core.sql

@@ -1,7 +1,110 @@
+-- TART baseline schema (squashed from historical migrations 001-024).
+--
+-- The old migration history created continuous aggregates with live refresh
+-- policies and later DROPped them (006, 015, 020). On a fresh database the
+-- policy jobs start running mid-sequence and DROP MATERIALIZED VIEW ... CASCADE
+-- deadlocks against the TimescaleDB job scheduler. This baseline creates only
+-- the final schema, so no aggregate is ever dropped during migration.
+--
+-- RULE FOR FUTURE MIGRATIONS: never DROP a continuous aggregate (or a table
+-- with policy jobs) that an EARLIER migration created — on a fresh database
+-- its background job may already be running and the DROP can deadlock with
+-- the scheduler. If a drop is unavoidable, remove the policies first and
+-- accept that the race still exists; prefer additive changes.
+--
+-- Column order note: several tables list columns in "historical" order
+-- (original columns first, later ALTER TABLE ADD COLUMNs last) so that a
+-- fresh database is catalog-identical to one that replayed the old history.
+
+-- Enable extensions
+CREATE EXTENSION IF NOT EXISTS timescaledb;
+CREATE EXTENSION IF NOT EXISTS pg_stat_statements;
+
+-- ============================================================
+-- Nodes table (regular PostgreSQL table - low cardinality, ~1024 rows)
+-- ============================================================
+CREATE TABLE IF NOT EXISTS nodes (
+    node_id TEXT PRIMARY KEY,
+    peer_id TEXT NOT NULL,
+    implementation_name TEXT NOT NULL,
+    implementation_version TEXT NOT NULL,
+    node_info JSONB NOT NULL,
+    connected_at TIMESTAMPTZ NOT NULL,
+    disconnected_at TIMESTAMPTZ,
+    last_seen_at TIMESTAMPTZ NOT NULL,
+    is_connected BOOLEAN DEFAULT true,
+    event_count BIGINT DEFAULT 0,
+    total_events BIGINT DEFAULT 0,
+    address TEXT,
+    created_at TIMESTAMPTZ DEFAULT CURRENT_TIMESTAMP
+);
+
+CREATE INDEX IF NOT EXISTS idx_nodes_connected ON nodes(is_connected, last_seen_at DESC) WHERE is_connected = true;
+CREATE INDEX IF NOT EXISTS idx_nodes_last_seen ON nodes(last_seen_at DESC);
+
+-- ============================================================
+-- Raw events hypertable: 1h browsing store with hot columns.
+-- All 115 event types are written here; aggregation lives in count tables.
+-- Historically created as 'events' and renamed; index names keep the
+-- original idx_events_* prefix on purpose.
+-- ============================================================
+CREATE TABLE IF NOT EXISTS ingested_raw_events (
+    timestamp    TIMESTAMPTZ NOT NULL,
+    node_id      TEXT        NOT NULL,
+    event_id     BIGINT      NOT NULL,
+    event_type   SMALLINT    NOT NULL,
+    data         JSONB       NOT NULL,
+    created_at   TIMESTAMPTZ NOT NULL DEFAULT NOW(),
+    -- Hot columns: frequently-queried JSONB fields promoted to real columns
+    slot         INT,
+    core         SMALLINT,
+    submission_id BIGINT
+);
+
+-- Convert to hypertable with 1-hour chunks
+SELECT create_hypertable('ingested_raw_events', 'timestamp',
+    chunk_time_interval => INTERVAL '1 hour',
+    create_default_indexes => FALSE,
+    if_not_exists => TRUE
+);
+
+-- Space partitioning on node_id (32 hash buckets for write distribution)
+SELECT add_dimension('ingested_raw_events', by_hash('node_id', 32), if_not_exists => TRUE);
+
+-- Minimal indexes (each index costs write throughput at 3M events/s)
+CREATE INDEX IF NOT EXISTS idx_events_node_time ON ingested_raw_events (node_id, timestamp DESC);
+CREATE INDEX IF NOT EXISTS idx_events_type_time ON ingested_raw_events (event_type, timestamp DESC);
+CREATE INDEX IF NOT EXISTS idx_events_slot ON ingested_raw_events (slot, timestamp DESC) WHERE slot IS NOT NULL;
+CREATE INDEX IF NOT EXISTS idx_events_core ON ingested_raw_events (core, timestamp DESC) WHERE core IS NOT NULL;
+CREATE INDEX IF NOT EXISTS idx_events_submission_id ON ingested_raw_events (node_id, submission_id, timestamp DESC) WHERE submission_id IS NOT NULL;
+
+-- Pure browsing store — aggressive 1h retention, checked every 5 minutes
+SELECT add_retention_policy('ingested_raw_events', INTERVAL '1 hour', schedule_interval => INTERVAL '5 minutes');
+
+-- VIEW alias: legacy endpoints reference 'events'.
+-- Created BEFORE the wp_hash column is added so the view's column list
+-- (SELECT * expands at creation time) matches the historical schema.
+CREATE VIEW events AS SELECT * FROM ingested_raw_events;
+
+-- wp_hash hot column (added after the view on purpose — see above)
+ALTER TABLE ingested_raw_events ADD COLUMN IF NOT EXISTS wp_hash BYTEA;
+CREATE INDEX IF NOT EXISTS idx_ire_wp_hash
+    ON ingested_raw_events (wp_hash, timestamp DESC) WHERE wp_hash IS NOT NULL;
+
+-- ============================================================
+-- Stats cache table for pre-computed aggregations
+-- ============================================================
+CREATE TABLE IF NOT EXISTS stats_cache (
+    key TEXT PRIMARY KEY,
+    value JSONB NOT NULL,
+    updated_at TIMESTAMPTZ DEFAULT CURRENT_TIMESTAMP
+);
+
+-- ============================================================
 -- Event type lookup table for human-readable names and grouping.
 -- TODO: Generate this table automatically from src/events.rs definitions.
 -- For now, this is a hardcoded list matching the JIP-3 telemetry protocol event types.
-
+-- ============================================================
 CREATE TABLE IF NOT EXISTS event_types (
     id SMALLINT PRIMARY KEY,
     name TEXT NOT NULL,
@@ -134,18 +237,3 @@ INSERT INTO event_types (id, name, group_name) VALUES
 (198, 'PreimageTransferred', 'preimages'),
 (199, 'PreimageDiscarded', 'preimages')
 ON CONFLICT (id) DO UPDATE SET name = EXCLUDED.name, group_name = EXCLUDED.group_name;
-
--- Convenience view: events pre-joined with type names and groups.
--- Use this in Grafana panels instead of raw events + JOIN.
-CREATE OR REPLACE VIEW events_view AS
-SELECT
-    e.timestamp,
-    e.node_id,
-    e.event_id,
-    e.event_type,
-    et.name AS event_name,
-    et.group_name AS event_group,
-    e.data,
-    e.created_at
-FROM events e
-LEFT JOIN event_types et ON e.event_type = et.id;