fix(alerts): a rule's page opens at once; its pickers read the latest values (#5473)

#5473 "Opening an alert rule takes up to 15 s: the Resource picker scans
every raw host-agent sample". The access log showed the edit page at 6-15 s
(one load gave up) while /alerts/ took 0.2 s. resources_for ran SELECT
DISTINCT over plugin_metrics, and Postgres has no skip scan, so it read every
host-agent sample; every rule page, source swap, rule save and MCP rule call
paid it. resources_for and metrics_for now read plugin_metric_latest, one
row per series, which lists the same names. A test fails if either reads
plugin_metrics again.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

steward/alerts/service.py

@@ -14,7 +14,7 @@ from sqlalchemy import func, or_, select
 from steward.models.alerts import (
     AlertEvent, AlertOperator, AlertRule, AlertState, AlertStateEnum, MaintenanceWindow,
 )
-from steward.models.metrics import PluginMetric, PluginMetricLatest
+from steward.models.metrics import PluginMetricLatest
 from steward.models.monitors import MONITOR_TYPE_VALUES, Monitor
 
 logger = logging.getLogger(__name__)
@@ -338,14 +338,17 @@ async def list_maintenance_windows(db) -> list[MaintenanceWindow]:
 async def resources_for(db, source_module: str) -> list[str]:
     """Sorted distinct resource names for a source module.
 
+    Read from the latest-value table, one row per series: a DISTINCT over
+    plugin_metrics reads every raw sample (Postgres has no skip scan), which
+    held the rule page for up to 15 s on host_agent (#5208's trap again).
+
     For monitor types (icmp/tcp/dns/http), supplements with monitor names so
     the list is populated even before any metrics have been recorded.
     """
     result = await db.execute(
-        select(PluginMetric.resource_name)
-        .where(PluginMetric.source_module == source_module)
+        select(PluginMetricLatest.resource_name)
+        .where(PluginMetricLatest.source_module == source_module)
         .distinct()
-        .order_by(PluginMetric.resource_name)
     )
     resources: set[str] = {r for (r,) in result}
     if source_module in MONITOR_TYPE_VALUES:
@@ -361,10 +364,10 @@ async def metrics_for(db, source_module: str) -> list[str]:
     metrics = metric_catalog().get(source_module, [])
     if source_module and not metrics:
         result = await db.execute(
-            select(PluginMetric.metric_name)
-            .where(PluginMetric.source_module == source_module)
+            select(PluginMetricLatest.metric_name)
+            .where(PluginMetricLatest.source_module == source_module)
             .distinct()
-            .order_by(PluginMetric.metric_name))
+            .order_by(PluginMetricLatest.metric_name))
         metrics = [r for (r,) in result]
     return metrics
 

tests/core/test_alert_pickers.py

@@ -0,0 +1,27 @@
+"""The rule form's Resource and Metric pickers read the latest-value table,
+one row per series, never the raw history: a DISTINCT over plugin_metrics
+reads every sample and held the rule page for up to 15 s on host_agent."""
+from types import SimpleNamespace
+
+from steward.alerts import service as alert_svc
+
+
+async def _sql(fn, source) -> list[str]:
+    """The SQL `fn` sends for `source`, against a session that returns no rows."""
+    sent: list[str] = []
+
+    async def execute(stmt):
+        sent.append(str(stmt.compile(compile_kwargs={"literal_binds": True})))
+        return []
+    await fn(SimpleNamespace(execute=execute), source)
+    return sent
+
+
+async def test_resources_come_from_the_latest_value_table():
+    (sql,) = await _sql(alert_svc.resources_for, "host_agent")
+    assert "FROM plugin_metric_latest" in sql and "plugin_metrics" not in sql
+
+
+async def test_a_dynamic_source_s_metrics_come_from_the_latest_value_table():
+    (sql,) = await _sql(alert_svc.metrics_for, "snmp")
+    assert "FROM plugin_metric_latest" in sql and "plugin_metrics" not in sql