(bugfix): coordination table provisioner fixes

CHANGELOG.md

@@ -4,6 +4,18 @@ All notable changes to this project will be documented in this file.
 
 The format is based on [Keep a Changelog](https://keepachangelog.com/en/1.1.0/), and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0.html).
 
+## [3.3.13] - 2026-04-26
+
+### Fixed
+
+- **HA lease / `42P01`:** Added `CoordinationTableProvisioner` with explicit `public.*` DDL; `InitializationHostedService` calls it immediately before bootstrap lease acquire (idempotent, same as post-migrate repair). `RuntimeLeaseServiceNpgsql` now uses `public.app_runtime_leases` in SQL so a non-default `search_path` cannot miss the table. Post-migrate verification requires `public.app_runtime_leases` plus `users` or `"VersionInfo"`.
+
+### Upgrade notes (Kubernetes / Helm)
+
+- **Pin container tags to the app semver** (e.g. `3.3.13` for server, client, reverseproxy) via `global.image.tag` and/or `components.*.image.tag`. The chart resolves the effective tag with `global.image.tag` when set (see `src/helm/templates/_helpers.tpl`).
+- **Do not rely on `latest` + `imagePullPolicy: IfNotPresent` alone** — nodes keep the first pulled digest, so you can run an old server binary while the OCI chart is already `3.3.13`. Use an explicit semver tag and/or `pullPolicy: Always` (or bump `global.rolloutNonce` / `global.rollme` per chart NOTES) when upgrading.
+- **Push all three images** for the tag you pin (`certs-ui/server`, `certs-ui/client`, `certs-ui/reverseproxy`) so every deployment can pull successfully.
+
 ## [3.3.12] - 2026-04-26
 
 ### Fixed

src/MaksIT.CertsUI.Engine/Infrastructure/CoordinationTableProvisioner.cs

@@ -0,0 +1,38 @@
+using Npgsql;
+
+namespace MaksIT.CertsUI.Engine.Infrastructure;
+
+/// <summary>
+/// Idempotent DDL for HA coordination tables in schema <c>public</c> (same shape as the AcmeChallengesAndRuntimeLeases migration). Used after FluentMigrator and again before bootstrap lease
+/// so <see cref="RuntimeLeaseServiceNpgsql"/> never runs against a missing <c>app_runtime_leases</c>.
+/// </summary>
+public static class CoordinationTableProvisioner {
+
+  /// <summary>Creates <c>public.acme_http_challenges</c> and <c>public.app_runtime_leases</c> if missing.</summary>
+  public static async Task EnsureAsync(string? connectionString, CancellationToken cancellationToken = default) {
+    if (string.IsNullOrWhiteSpace(connectionString))
+      return;
+
+    await using var conn = new NpgsqlConnection(connectionString);
+    await conn.OpenAsync(cancellationToken).ConfigureAwait(false);
+
+    await using var cmd = new NpgsqlCommand(
+      """
+      CREATE TABLE IF NOT EXISTS public.acme_http_challenges (
+        file_name text NOT NULL PRIMARY KEY,
+        token_value text NOT NULL,
+        created_at_utc timestamp with time zone NOT NULL
+      );
+      CREATE INDEX IF NOT EXISTS "IX_acme_http_challenges_created_at_utc" ON public.acme_http_challenges (created_at_utc);
+      CREATE TABLE IF NOT EXISTS public.app_runtime_leases (
+        lease_name text NOT NULL PRIMARY KEY,
+        holder_id text NOT NULL,
+        version bigint NOT NULL DEFAULT 1,
+        acquired_at_utc timestamp with time zone NOT NULL,
+        expires_at_utc timestamp with time zone NOT NULL
+      );
+      """,
+      conn);
+    await cmd.ExecuteNonQueryAsync(cancellationToken).ConfigureAwait(false);
+  }
+}

src/MaksIT.CertsUI.Engine/Infrastructure/RunMigrationsService.cs

@@ -38,7 +38,7 @@ public sealed class RunMigrationsService(
     await EnsureDatabaseExistsAsync(cancellationToken).ConfigureAwait(false);
     await BaselineExistingEfDatabaseAsync(cancellationToken).ConfigureAwait(false);
     await Task.Run(() => migrationRunner.MigrateUp(), cancellationToken).ConfigureAwait(false);
-    await EnsureCoordinationTablesAsync(cancellationToken).ConfigureAwait(false);
+    await CoordinationTableProvisioner.EnsureAsync(config.ConnectionString, cancellationToken).ConfigureAwait(false);
     await VerifyCoreSchemaAsync(cancellationToken).ConfigureAwait(false);
     logger.LogInformation("Certs database migrations completed.");
   }
@@ -50,12 +50,12 @@ public sealed class RunMigrationsService(
 
     await using var cmd = new NpgsqlCommand(
       """
-      SELECT EXISTS (
-        SELECT 1 FROM information_schema.tables
-        WHERE table_schema = 'public' AND table_name = 'users')
-      OR EXISTS (
-        SELECT 1 FROM information_schema.tables
-        WHERE table_schema = 'public' AND table_name = 'VersionInfo');
+      SELECT
+        EXISTS (SELECT 1 FROM information_schema.tables WHERE table_schema = 'public' AND table_name = 'app_runtime_leases')
+        AND (
+          EXISTS (SELECT 1 FROM information_schema.tables WHERE table_schema = 'public' AND table_name = 'users')
+          OR EXISTS (SELECT 1 FROM information_schema.tables WHERE table_schema = 'public' AND table_name = 'VersionInfo')
+        );
       """,
       conn);
 
@@ -64,37 +64,8 @@ public sealed class RunMigrationsService(
       return;
 
     throw new InvalidOperationException(
-      "After FluentMigrator MigrateUp(), the target database still has no \"users\" or \"VersionInfo\" table in schema \"public\". " +
-      "Confirm the connection string Database= value, that the role can CREATE TABLE, and that FluentMigrator is not in preview/connectionless mode (non-empty connection string).");
-  }
-
-  /// <summary>
-  /// Idempotent DDL for HA tables from <see cref="AcmeChallengesAndRuntimeLeases"/>.
-  /// When <c>VersionInfo</c> already lists that migration but the tables are missing (restore drift, partial apply),
-  /// FluentMigrator will not re-run <c>Up()</c>; this repair keeps lease and HTTP-01 persistence working.
-  /// </summary>
-  private async Task EnsureCoordinationTablesAsync(CancellationToken cancellationToken) {
-    await using var conn = new NpgsqlConnection(config.ConnectionString);
-    await conn.OpenAsync(cancellationToken).ConfigureAwait(false);
-
-    await using var cmd = new NpgsqlCommand(
-      """
-      CREATE TABLE IF NOT EXISTS acme_http_challenges (
-        file_name text NOT NULL PRIMARY KEY,
-        token_value text NOT NULL,
-        created_at_utc timestamp with time zone NOT NULL
-      );
-      CREATE INDEX IF NOT EXISTS "IX_acme_http_challenges_created_at_utc" ON acme_http_challenges (created_at_utc);
-      CREATE TABLE IF NOT EXISTS app_runtime_leases (
-        lease_name text NOT NULL PRIMARY KEY,
-        holder_id text NOT NULL,
-        version bigint NOT NULL DEFAULT 1,
-        acquired_at_utc timestamp with time zone NOT NULL,
-        expires_at_utc timestamp with time zone NOT NULL
-      );
-      """,
-      conn);
-    await cmd.ExecuteNonQueryAsync(cancellationToken).ConfigureAwait(false);
+      "After migrations and coordination DDL, schema \"public\" is missing \"app_runtime_leases\" and/or core tables (\"users\" / \"VersionInfo\"). " +
+      "Confirm Database= in the connection string, role CREATE privileges, and that FluentMigrator committed (non-empty connection string).");
   }
 
   private async Task EnsureDatabaseExistsAsync(CancellationToken cancellationToken) {

src/MaksIT.CertsUI.Engine/Infrastructure/RuntimeLeaseServiceNpgsql.cs

@@ -29,15 +29,15 @@ public sealed class RuntimeLeaseServiceNpgsql(
 
       await using var cmd = new NpgsqlCommand(
         """
-        INSERT INTO app_runtime_leases (lease_name, holder_id, version, acquired_at_utc, expires_at_utc)
+        INSERT INTO public.app_runtime_leases (lease_name, holder_id, version, acquired_at_utc, expires_at_utc)
         VALUES (@name, @holder, 1, @acquired, @expires)
         ON CONFLICT (lease_name) DO UPDATE
         SET holder_id = EXCLUDED.holder_id,
-            version = app_runtime_leases.version + 1,
+            version = public.app_runtime_leases.version + 1,
             acquired_at_utc = EXCLUDED.acquired_at_utc,
             expires_at_utc = EXCLUDED.expires_at_utc
-        WHERE app_runtime_leases.expires_at_utc < EXCLUDED.acquired_at_utc
-           OR app_runtime_leases.holder_id = EXCLUDED.holder_id
+        WHERE public.app_runtime_leases.expires_at_utc < EXCLUDED.acquired_at_utc
+           OR public.app_runtime_leases.holder_id = EXCLUDED.holder_id
         RETURNING holder_id;
         """,
         conn);
@@ -72,7 +72,7 @@ public sealed class RuntimeLeaseServiceNpgsql(
 
       await using var cmd = new NpgsqlCommand(
         """
-        DELETE FROM app_runtime_leases
+        DELETE FROM public.app_runtime_leases
         WHERE lease_name = @name AND holder_id = @holder;
         """,
         conn);

src/MaksIT.CertsUI/HostedServices/InitializationHostedService.cs

@@ -1,4 +1,5 @@
 using Microsoft.Extensions.Options;
+using MaksIT.CertsUI.Engine;
 using MaksIT.CertsUI.Engine.DomainServices;
 using MaksIT.CertsUI.Engine.Infrastructure;
 using MaksIT.CertsUI.Engine.RuntimeCoordination;
@@ -7,8 +8,8 @@ namespace MaksIT.CertsUI.HostedServices;
 
 /// <summary>
 /// Runs identity bootstrap before the API starts serving requests. FluentMigrator already ran in <c>Program.cs</c>
-/// before the host starts. The bootstrap lease ensures only one replica writes against shared
-/// <see cref="Configuration.CertsUIEngineConfiguration.DataFolder"/>.
+/// before the host starts; coordination tables in <c>public</c> are ensured again here before the bootstrap lease.
+/// The bootstrap lease ensures only one replica writes against shared <see cref="Configuration.CertsUIEngineConfiguration.DataFolder"/>.
 /// </summary>
 public sealed class InitializationHostedService(
   ILogger<InitializationHostedService> logger,
@@ -27,6 +28,9 @@ public sealed class InitializationHostedService(
       try {
         logger.LogInformation("Running startup initialization...");
 
+        var engineConfig = serviceProvider.GetRequiredService<ICertsEngineConfiguration>();
+        await CoordinationTableProvisioner.EnsureAsync(engineConfig.ConnectionString, cancellationToken).ConfigureAwait(false);
+
         var holder = runtimeInstance.InstanceId;
         var acquired = await runtimeLease.TryAcquireAsync(RuntimeLeaseNames.Bootstrap, holder, BootstrapLeaseTtl, cancellationToken).ConfigureAwait(false);
         if (!acquired.IsSuccess)

src/MaksIT.CertsUI/MaksIT.CertsUI.csproj

@@ -1,7 +1,7 @@
 <Project Sdk="Microsoft.NET.Sdk.Web">
 
   <PropertyGroup>
-    <Version>3.3.12</Version>
+    <Version>3.3.13</Version>
     <TargetFramework>net10.0</TargetFramework>
     <Nullable>enable</Nullable>
     <ImplicitUsings>enable</ImplicitUsings>

src/helm/values.yaml

@@ -1,8 +1,8 @@
 global:
   imagePullSecrets: []
   image:
-    # Uncomment to override every component (global wins when set). Otherwise use each components.*.image.
-    # No Chart.appVersion.
+    # When non-empty, overrides every components.*.image.tag (see _helpers.tpl). Production/staging: pin to released
+    # semver (e.g. 3.3.13) and use pullPolicy Always or bump tag each release — do not rely on :latest + IfNotPresent alone.
     # tag: "latest"
     # pullPolicy: IfNotPresent
   # Optional rollout tuning (see NOTES): pin a fixed pod annotation or add a nonce for frozen/git-rendered manifests.