V16: Cache Version Mechanism (#19747)

* Add RepositoryCacheVersion table

* Add repository

* Add Cache version lock

* Add GetAll method to repository

* Add RepositoryCacheVersionService

* Remember to add lock in data creator

* Work my way out of constructor hell

This is why we use DI folks. 🤦

* Add checks to specific cache policies

* Fix migration

* Add to schema creator

* Fix database access

* Initialize the cache version on in memory miss

* Make cache version service internal

* Add tests

* Apply suggestions from code review

Co-authored-by: Andy Butland <[email protected]>

* Add missing obsoletions

* Prefer full name

---------

Co-authored-by: Andy Butland <[email protected]>

Author: nikolajlauridsen

src/Umbraco.Core/Cache/IRepositoryCacheVersionService.cs

@@ -0,0 +1,28 @@
+namespace Umbraco.Cms.Core.Cache;
+
+/// <summary>
+/// Provides methods to manage and validate cache versioning for repository entities,
+/// ensuring cache consistency with the underlying database.
+/// </summary>
+public interface IRepositoryCacheVersionService
+{
+    /// <summary>
+    /// Validates if the cache is synced with the database.
+    /// </summary>
+    /// <typeparam name="TEntity">The type of the cached entity.</typeparam>
+    /// <returns>True if cache is synced, false if cache needs fast-forwarding.</returns>
+    Task<bool> IsCacheSyncedAsync<TEntity>()
+        where TEntity : class;
+
+    /// <summary>
+    /// Registers a cache update for the specified entity type.
+    /// </summary>
+    /// <typeparam name="TEntity">The type of the cached entity.</typeparam>
+    Task SetCacheUpdatedAsync<TEntity>()
+        where TEntity : class;
+
+    /// <summary>
+    /// Registers that the cache has been synced with the database.
+    /// </summary>
+    Task SetCachesSyncedAsync();
+}

src/Umbraco.Core/Cache/RepositoryCacheVersionService.cs

@@ -0,0 +1,118 @@
+using System.Collections.Concurrent;
+using Microsoft.Extensions.Logging;
+using Umbraco.Cms.Core.Models;
+using Umbraco.Cms.Core.Persistence.Repositories;
+using Umbraco.Cms.Core.Scoping;
+
+namespace Umbraco.Cms.Core.Cache;
+
+/// <inheritdoc />
+internal class RepositoryCacheVersionService : IRepositoryCacheVersionService
+{
+    private readonly ICoreScopeProvider _scopeProvider;
+    private readonly IRepositoryCacheVersionRepository _repositoryCacheVersionRepository;
+    private readonly ILogger<RepositoryCacheVersionService> _logger;
+    private readonly ConcurrentDictionary<string, Guid> _cacheVersions = new();
+
+    public RepositoryCacheVersionService(
+        ICoreScopeProvider scopeProvider,
+        IRepositoryCacheVersionRepository repositoryCacheVersionRepository,
+        ILogger<RepositoryCacheVersionService> logger)
+    {
+        _scopeProvider = scopeProvider;
+        _repositoryCacheVersionRepository = repositoryCacheVersionRepository;
+        _logger = logger;
+    }
+
+    /// <inheritdoc />
+    public async Task<bool> IsCacheSyncedAsync<TEntity>()
+        where TEntity : class
+    {
+        _logger.LogDebug("Checking if cache for {EntityType} is synced", typeof(TEntity).Name);
+
+        // We have to take a read lock to ensure the cache is not being updated while we check the version.
+        using ICoreScope scope = _scopeProvider.CreateCoreScope(autoComplete: true);
+        scope.ReadLock(Constants.Locks.CacheVersion);
+
+        var cacheKey = GetCacheKey<TEntity>();
+
+        RepositoryCacheVersion? databaseVersion = await _repositoryCacheVersionRepository.GetAsync(cacheKey);
+
+        if (databaseVersion?.Version is null)
+        {
+            _logger.LogDebug("Cache for {EntityType} has no version in the database, considering it synced", typeof(TEntity).Name);
+
+            // If the database version is null, it means the cache has never been initialized, so we consider it synced.
+            return true;
+        }
+
+        if (_cacheVersions.TryGetValue(cacheKey, out Guid localVersion) is false)
+        {
+            _logger.LogDebug("Cache for {EntityType} is not initialized, considering it synced", typeof(TEntity).Name);
+
+            // We're not initialized yet, so cache is empty, which means cache is synced.
+            // Since the cache is most likely no longer empty, we should set the cache version to the database version.
+            _cacheVersions[cacheKey] = Guid.Parse(databaseVersion.Version);
+            return true;
+        }
+
+        // We could've parsed this in the repository layer; however, the fact that we are using a Guid is an implementation detail.
+        if (localVersion != Guid.Parse(databaseVersion.Version))
+        {
+            _logger.LogDebug(
+                "Cache for {EntityType} is not synced: local version {LocalVersion} does not match database version {DatabaseVersion}",
+                typeof(TEntity).Name,
+                localVersion,
+                databaseVersion.Version);
+            return false;
+        }
+
+        _logger.LogDebug("Cache for {EntityType} is synced", typeof(TEntity).Name);
+        return true;
+    }
+
+    /// <inheritdoc />
+    public async Task SetCacheUpdatedAsync<TEntity>()
+        where TEntity : class
+    {
+        using ICoreScope scope = _scopeProvider.CreateCoreScope();
+
+        // We have to take a write lock to ensure the cache is not being read while we update the version.
+        scope.WriteLock(Constants.Locks.CacheVersion);
+
+        var cacheKey = GetCacheKey<TEntity>();
+        var newVersion = Guid.NewGuid();
+
+        _logger.LogDebug("Setting cache for {EntityType} to version {Version}", typeof(TEntity).Name, newVersion);
+        await _repositoryCacheVersionRepository.SaveAsync(new RepositoryCacheVersion { Identifier = cacheKey, Version = newVersion.ToString() });
+        _cacheVersions[cacheKey] = newVersion;
+
+        scope.Complete();
+    }
+
+    /// <inheritdoc />
+    public async Task SetCachesSyncedAsync()
+    {
+        using ICoreScope scope = _scopeProvider.CreateCoreScope();
+        scope.ReadLock(Constants.Locks.CacheVersion);
+
+        // We always sync all caches versions, so it's safe to assume all caches are synced at this point.
+        IEnumerable<RepositoryCacheVersion> cacheVersions = await _repositoryCacheVersionRepository.GetAllAsync();
+
+        foreach (RepositoryCacheVersion version in cacheVersions)
+        {
+            if (version.Version is null)
+            {
+                continue;
+            }
+
+            _cacheVersions[version.Identifier] = Guid.Parse(version.Version);
+        }
+
+        scope.Complete();
+    }
+
+    internal string GetCacheKey<TEntity>()
+        where TEntity : class =>
+        typeof(TEntity).FullName ?? typeof(TEntity).Name;
+}

src/Umbraco.Core/Cache/SingleServerCacheVersionService.cs

@@ -0,0 +1,23 @@
+namespace Umbraco.Cms.Core.Cache;
+
+/// <summary>
+/// A simple cache version service that assumes the cache is always in sync.
+/// <remarks>
+/// This is useful in scenarios where you have a single server setup and do not need to manage cache synchronization across multiple servers.
+/// </remarks>
+/// </summary>
+public class SingleServerCacheVersionService : IRepositoryCacheVersionService
+{
+    /// <inheritdoc />
+    public Task<bool> IsCacheSyncedAsync<TEntity>()
+        where TEntity : class
+        => Task.FromResult(true);
+
+    /// <inheritdoc />
+    public Task SetCacheUpdatedAsync<TEntity>()
+        where TEntity : class
+        => Task.CompletedTask;
+
+    /// <inheritdoc />
+    public Task SetCachesSyncedAsync() => Task.CompletedTask;
+}

src/Umbraco.Core/DependencyInjection/UmbracoBuilder.cs

@@ -341,6 +341,7 @@ private void AddCoreServices()
             Services.AddUnique<ILocalizedTextService>(factory => new LocalizedTextService(
                 factory.GetRequiredService<Lazy<LocalizedTextServiceFileSources>>(),
                 factory.GetRequiredService<ILogger<LocalizedTextService>>()));
+            Services.AddUnique<IRepositoryCacheVersionService, RepositoryCacheVersionService>();
 
             Services.AddUnique<IEntityXmlSerializer, EntityXmlSerializer>();
 

src/Umbraco.Core/Models/RepositoryCacheVersion.cs

@@ -0,0 +1,17 @@
+namespace Umbraco.Cms.Core.Models;
+
+/// <summary>
+/// Represents a version of a repository cache.
+/// </summary>
+public class RepositoryCacheVersion
+{
+    /// <summary>
+    /// The unique identifier for the cache.
+    /// </summary>
+    public required string Identifier { get; init; }
+
+    /// <summary>
+    /// The identifier of the version of the cache.
+    /// </summary>
+    public required string? Version { get; init; }
+}

src/Umbraco.Core/Persistence/Constants-DatabaseSchema.cs

@@ -98,6 +98,8 @@ public static class Tables
             public const string Webhook2Headers = Webhook + "2Headers";
             public const string WebhookLog = Webhook + "Log";
             public const string WebhookRequest = Webhook + "Request";
+
+            public const string RepositoryCacheVersion = TableNamePrefix + "RepositoryCacheVersion";
         }
     }
 }

src/Umbraco.Core/Persistence/Constants-Locks.cs

@@ -80,5 +80,10 @@ public static class Locks
         ///     All webhook logs.
         /// </summary>
         public const int WebhookLogs = -343;
+
+        /// <summary>
+        ///    The cache version.
+        /// </summary>
+        public const int CacheVersion = -346;
     }
 }

src/Umbraco.Core/Persistence/Repositories/IRepositoryCacheVersionRepository.cs

@@ -0,0 +1,32 @@
+using Umbraco.Cms.Core.Models;
+
+namespace Umbraco.Cms.Core.Persistence.Repositories;
+
+/// <summary>
+/// Defines methods for accessing and persisting <see cref="RepositoryCacheVersion"/> entities.
+/// </summary>
+public interface IRepositoryCacheVersionRepository : IRepository
+{
+    /// <summary>
+    /// Gets a <see cref="RepositoryCacheVersion"/> by its identifier.
+    /// </summary>
+    /// <param name="identifier">The unique identifier of the cache version.</param>
+    /// <returns>
+    /// A <see cref="RepositoryCacheVersion"/> if found; otherwise, <c>null</c>.
+    /// </returns>
+    Task<RepositoryCacheVersion?> GetAsync(string identifier);
+
+    /// <summary>
+    /// Gets all <see cref="RepositoryCacheVersion"/> entities.
+    /// </summary>
+    /// <returns>
+    /// An <see cref="IEnumerable{RepositoryCacheVersion}"/> containing all cache versions.
+    /// </returns>
+    Task<IEnumerable<RepositoryCacheVersion>> GetAllAsync();
+
+    /// <summary>
+    /// Saves the specified <see cref="RepositoryCacheVersion"/>.
+    /// </summary>
+    /// <param name="repositoryCacheVersion">The cache version entity to save.</param>
+    Task SaveAsync(RepositoryCacheVersion repositoryCacheVersion);
+}

src/Umbraco.Infrastructure/Cache/DefaultRepositoryCachePolicy.cs

@@ -1,6 +1,8 @@
 // Copyright (c) Umbraco.
 // See LICENSE for more details.
 
+using Microsoft.Extensions.DependencyInjection;
+using Umbraco.Cms.Core.DependencyInjection;
 using Umbraco.Cms.Core.Models.Entities;
 using Umbraco.Cms.Infrastructure.Scoping;
 using Umbraco.Extensions;
@@ -24,10 +26,27 @@ public class DefaultRepositoryCachePolicy<TEntity, TId> : RepositoryCachePolicyB
     private static readonly TEntity[] _emptyEntities = new TEntity[0]; // const
     private readonly RepositoryCachePolicyOptions _options;
 
-    public DefaultRepositoryCachePolicy(IAppPolicyCache cache, IScopeAccessor scopeAccessor, RepositoryCachePolicyOptions options)
-        : base(cache, scopeAccessor) =>
+    public DefaultRepositoryCachePolicy(
+        IAppPolicyCache cache,
+        IScopeAccessor scopeAccessor,
+        RepositoryCachePolicyOptions options,
+        IRepositoryCacheVersionService repositoryCacheVersionService)
+        : base(cache, scopeAccessor, repositoryCacheVersionService) =>
         _options = options ?? throw new ArgumentNullException(nameof(options));
 
+    [Obsolete("Please use the constructor with all parameters. Scheduled for removal in Umbraco 18.")]
+    public DefaultRepositoryCachePolicy(
+        IAppPolicyCache cache,
+        IScopeAccessor scopeAccessor,
+        RepositoryCachePolicyOptions options)
+        : this(
+            cache,
+            scopeAccessor,
+            options,
+            StaticServiceProvider.Instance.GetRequiredService<IRepositoryCacheVersionService>())
+    {
+    }
+
     protected string EntityTypeCacheKey { get; } = $"uRepo_{typeof(TEntity).Name}_";
 
     /// <inheritdoc />
@@ -98,6 +117,10 @@ public override void Update(TEntity entity, Action<TEntity> persistUpdated)
 
             throw;
         }
+
+        // We've changed the entity, register cache change for other servers.
+        // We assume that if something goes wrong, we'll roll back, so don't need to register the change.
+        RegisterCacheChange();
     }
 
     /// <inheritdoc />
@@ -122,11 +145,16 @@ public override void Delete(TEntity entity, Action<TEntity> persistDeleted)
             // if there's a GetAllCacheAllowZeroCount cache, ensure it is cleared
             Cache.Clear(EntityTypeCacheKey);
         }
+
+        // We've removed an entity, register cache change for other servers.
+        RegisterCacheChange();
     }
 
     /// <inheritdoc />
     public override TEntity? Get(TId? id, Func<TId?, TEntity?> performGet, Func<TId[]?, IEnumerable<TEntity>?> performGetAll)
     {
+        EnsureCacheIsSynced();
+
         var cacheKey = GetEntityCacheKey(id);
 
         TEntity? fromCache = Cache.GetCacheItem<TEntity>(cacheKey);
@@ -163,13 +191,15 @@ public override void Delete(TEntity entity, Action<TEntity> persistDeleted)
     /// <inheritdoc />
     public override TEntity? GetCached(TId id)
     {
+        EnsureCacheIsSynced();
         var cacheKey = GetEntityCacheKey(id);
         return Cache.GetCacheItem<TEntity>(cacheKey);
     }
 
     /// <inheritdoc />
     public override bool Exists(TId id, Func<TId, bool> performExists, Func<TId[], IEnumerable<TEntity>?> performGetAll)
     {
+        EnsureCacheIsSynced();
         // if found in cache the return else check
         var cacheKey = GetEntityCacheKey(id);
         TEntity? fromCache = Cache.GetCacheItem<TEntity>(cacheKey);
@@ -179,6 +209,7 @@ public override bool Exists(TId id, Func<TId, bool> performExists, Func<TId[], I
     /// <inheritdoc />
     public override TEntity[] GetAll(TId[]? ids, Func<TId[]?, IEnumerable<TEntity>?> performGetAll)
     {
+        EnsureCacheIsSynced();
         if (ids?.Length > 0)
         {
             // try to get each entity from the cache

src/Umbraco.Infrastructure/Cache/FullDataSetRepositoryCachePolicy.cs

@@ -100,6 +100,10 @@ public override void Update(TEntity entity, Action<TEntity> persistUpdated)
         {
             ClearAll();
         }
+
+        // We've changed the entity, register cache change for other servers.
+        // We assume that if something goes wrong, we'll roll back, so don't need to register the change.
+        RegisterCacheChange();
     }
 
     /// <inheritdoc />
@@ -118,11 +122,17 @@ public override void Delete(TEntity entity, Action<TEntity> persistDeleted)
         {
             ClearAll();
         }
+
+        // We've changed the entity, register cache change for other servers.
+        // We assume that if something goes wrong, we'll roll back, so don't need to register the change.
+        RegisterCacheChange();
     }
 
     /// <inheritdoc />
     public override TEntity? Get(TId? id, Func<TId?, TEntity?> performGet, Func<TId[]?, IEnumerable<TEntity>?> performGetAll)
     {
+        EnsureCacheIsSynced();
+
         // get all from the cache, then look for the entity
         IEnumerable<TEntity> all = GetAllCached(performGetAll);
         TEntity? entity = all.FirstOrDefault(x => _entityGetId(x)?.Equals(id) ?? false);
@@ -135,6 +145,8 @@ public override void Delete(TEntity entity, Action<TEntity> persistDeleted)
     /// <inheritdoc />
     public override TEntity? GetCached(TId id)
     {
+        EnsureCacheIsSynced();
+
         // get all from the cache -- and only the cache, then look for the entity
         DeepCloneableList<TEntity>? all = Cache.GetCacheItem<DeepCloneableList<TEntity>>(GetEntityTypeCacheKey());
         TEntity? entity = all?.FirstOrDefault(x => _entityGetId(x)?.Equals(id) ?? false);
@@ -147,6 +159,8 @@ public override void Delete(TEntity entity, Action<TEntity> persistDeleted)
     /// <inheritdoc />
     public override bool Exists(TId id, Func<TId, bool> performExits, Func<TId[], IEnumerable<TEntity>?> performGetAll)
     {
+        EnsureCacheIsSynced();
+
         // get all as one set, then look for the entity
         IEnumerable<TEntity> all = GetAllCached(performGetAll);
         return all.Any(x => _entityGetId(x)?.Equals(id) ?? false);
@@ -155,6 +169,8 @@ public override bool Exists(TId id, Func<TId, bool> performExits, Func<TId[], IE
     /// <inheritdoc />
     public override TEntity[] GetAll(TId[]? ids, Func<TId[], IEnumerable<TEntity>?> performGetAll)
     {
+        EnsureCacheIsSynced();
+
         // get all as one set, from cache if possible, else repo
         IEnumerable<TEntity> all = GetAllCached(performGetAll);