railwatch 0.8.4 → 0.8.5

checksums.yaml

@@ -1,7 +1,7 @@
 ---
 SHA256:
-  metadata.gz: f52b89a567e0ac0a05281db9b77499961ee7abf9cbfbf663070600d8885818a3
-  data.tar.gz: 05cdb2e7d19bf73edb2dd80550b1a59bf39c83ad24b951c7f7d43b3ddee9a850
+  metadata.gz: ec770e9d790b50b269a6bfde1e37d453ff6deb312216ab05d8106a50bb72d9b2
+  data.tar.gz: 9b9bff09f0821fefd39ae31d7c2ac826b3a6044d899aaf53a4e8e712a093d8f2
 SHA512:
-  metadata.gz: d3d4c4f9698de7d6f522d4c171962e959efccd4435ff07e036505ced415aaa43e4b134df949f18c61952c4b22738d7074f49afcc1209cf1088fe41ea6cae564b
-  data.tar.gz: 8f4c96092fc3a62f3538667c4d4f2e40de420e9f2ccacf7305b0b4b4fd99ca9b098cb830cced5b9def8a618d3d1e16e24e478d4461f2599e0047121b79e2df2e
+  metadata.gz: eb2fc4d6a744ca775895c0bbcb61b15321df7c3be1669c42c8e83b36563ff76fdd2c4aad2308af5eeeaa17bba68833e61a8e7c9c726adcc5b0199b502ee68c5b
+  data.tar.gz: e0685c8d6b3b9564728239e1cb6978e4eaaf888797fdf000deff1fbfcf642849424baca6e11d0d2b2d58afc40cebe23d20b14415e802835fe294dc5c2aa55e41

data/CHANGELOG.md

@@ -6,6 +6,15 @@
      filled in by the release commit, which is also the only commit that
      touches lib/railwatch/version.rb and Gemfile.lock. See CONTRIBUTING.md. -->
 
+## 0.8.5 (2026-09-25)
+
+- A new migration adds a partial covering index for the per-tenant sums
+  on the Tenants page and in MCP's `list_tenants`. Over 30 days of
+  rebulk-system's 358,000 tagged requests they took 8.1 s, reading every
+  row from the table; they now take 131 ms. It holds only tagged rows,
+  so an app without tenants gets an empty index. It took 1.6 s to build
+  on a 16 GB file.
+
 ## 0.8.4 (2026-09-23)
 
 - A `class:` filter for a job class with many attempts in the window

data/app/models/railwatch/telemetry/tenant.rb

@@ -120,8 +120,18 @@ module Railwatch
 
       # -- Aggregation steps -----------------------------------------------------
 
+      # The per-tenant sums read through idx_executions_tenant_summary, which
+      # holds every column they need for tagged rows only. SQLite prefers the
+      # narrower (app_tenant, occurred_at) index on its own estimate and then
+      # fetches every row from the table (8 s against 131 ms over 30 days).
+      def self.summary_scope(kind)
+        scope = Telemetry::Execution
+        scope = scope.from("#{scope.quoted_table_name} INDEXED BY idx_executions_tenant_summary") if TelemetryRecord.sqlite?
+        scope.where(kind: kind).where.not(app_tenant: nil)
+      end
+
       def self.absorb_requests(rows, from, to, q)
-        filtered(Telemetry::Execution.requests.between(from, to), q).group(:app_tenant)
+        filtered(summary_scope("request").between(from, to), q).group(:app_tenant)
           .pluck(:app_tenant, COUNT_SQL, ERRORS_SQL, Arel.sql("AVG(duration)"), Arel.sql("MAX(duration)"), USERS_SQL)
           .each do |tenant, count, errors, avg, max, users|
             rows[tenant].merge!(requests: count, errors: errors, avg: ms(avg), max: ms(max), users: users)
@@ -129,7 +139,7 @@ module Railwatch
       end
 
       def self.absorb_jobs(rows, from, to, q)
-        filtered(Telemetry::Execution.jobs.between(from, to), q).group(:app_tenant)
+        filtered(summary_scope("job_attempt").between(from, to), q).group(:app_tenant)
           .pluck(:app_tenant, COUNT_SQL, FAILED_SQL)
           .each { |tenant, count, failed| rows[tenant].merge!(jobs: count, failed_jobs: failed) }
       end
@@ -139,7 +149,7 @@ module Railwatch
       end
 
       def self.absorb_last_seen(rows, from, to, q)
-        filtered(Telemetry::Execution.between(from, to), q).group(:app_tenant).maximum(:occurred_at)
+        filtered(summary_scope(Telemetry::Execution::KINDS).between(from, to), q).group(:app_tenant).maximum(:occurred_at)
           .each { |tenant, at| rows[tenant][:last_seen_at] = at }
       end
 

data/db/railwatch_telemetry_migrate/20260925000000_add_tenant_summary_index.rb

@@ -0,0 +1,17 @@
+# frozen_string_literal: true
+
+# The Tenants page and MCP's list_tenants sum every tenant-tagged request
+# and job in the window per tenant. Through (app_tenant, occurred_at)
+# SQLite finds the rows but fetches each from the table for kind, status,
+# outcome, duration and user_ref: 8.1 s over 30 days on rebulk-system
+# (358,000 tagged requests), and 24 s once in production, which held the
+# web process -- and every ingest request queued behind it -- the whole
+# time. This index carries every column those aggregates read, only for
+# tagged rows: 131 ms for the same query, 41 MB, 1.6 s to build on a copy
+# of that 16 GB file. An untagged app gets an empty index.
+class AddTenantSummaryIndex < ActiveRecord::Migration[8.1]
+  def change
+    add_index :executions, [ :kind, :app_tenant, :occurred_at, :status, :outcome, :duration, :user_ref ],
+              name: "idx_executions_tenant_summary", where: "app_tenant IS NOT NULL"
+  end
+end

data/lib/railwatch/version.rb

@@ -1,5 +1,5 @@
 # frozen_string_literal: true
 
 module Railwatch
-  VERSION = "0.8.4"
+  VERSION = "0.8.5"
 end

metadata

@@ -1,7 +1,7 @@
 --- !ruby/object:Gem::Specification
 name: railwatch
 version: !ruby/object:Gem::Version
-  version: 0.8.4
+  version: 0.8.5
 platform: ruby
 authors:
 - Cole Robertson
@@ -405,6 +405,7 @@ files:
 - db/railwatch_telemetry_migrate/20260919000100_create_export_queue.rb
 - db/railwatch_telemetry_migrate/20260922000000_add_covering_indexes_for_dashboard_aggregates.rb
 - db/railwatch_telemetry_migrate/20260923000000_add_people_index.rb
+- db/railwatch_telemetry_migrate/20260925000000_add_tenant_summary_index.rb
 - docs/ai-and-mcp.md
 - docs/configuration.md
 - docs/embedded.md