Sitecore and Redis lessons learned

I noticed that my previous post about Redis is one of the most popular on my blog. Since I’ve been using Redis for a while I decided to write another post with some of the lessons learned.

Sitecore connectivity to Redis

The first step in getting Redis to work with Sitecore is to ensure there is connectivity between them. When Sitecore starts up it will ping Redis. The Sitecore log will contain something like below when connectivity to Redis is established successfully. Notice the Redis response to the Ping and the message that the endpoint returned with success.

11056 11:06:22 INFO  Sending critical tracer: Interactive/jeroen.redis.cache.windows.net:6380
11056 11:06:22 INFO  Writing to Interactive/jeroen.redis.cache.windows.net:6380: ECHO
11056 11:06:22 INFO  Flushing outbound buffer
11056 11:06:22 INFO  Starting read
11056 11:06:22 INFO  Connect complete: jeroen.redis.cache.windows.net:6380
11056 11:06:22 INFO  Response from Interactive/jeroen.redis.cache.windows.net:6380 / ECHO: BulkString: 16 bytes
WIN-RCJOA5J2MOL:Write 11:06:22 INFO  Writing to Interactive/jeroen.redis.cache.windows.net:6380: GET __Booksleeve_TieBreak
WIN-RCJOA5J2MOL:Write 11:06:22 INFO  Writing to Interactive/jeroen.redis.cache.windows.net:6380: PING
8912 11:06:22 INFO  Response from Interactive/jeroen.redis.cache.windows.net:6380 / GET __Booksleeve_TieBreak: (null)
8912 11:06:22 INFO  Response from Interactive/jeroen.redis.cache.windows.net:6380 / PING: SimpleString: PONG
1068 11:06:22 INFO  All tasks completed cleanly, IOCP: (Busy=0,Free=800,Min=800,Max=800), WORKER: (Busy=43,Free=757,Min=789,Max=800)
1068 11:06:22 INFO  jeroen.redis.cache.windows.net:6380 returned with success

There can be a variety of issues which prevents Sitecore from connecting to Redis:

  • Wrong Redis engine version: Sitecore does not work with Redis engine version 4 or 5. This is easy to get wrong especially if using AWS ElastiCache which currently defaults to version 5.0.3. When using AWS ElastiCache make sure to select version 3.2.6. This issue is not obvious from the log. When using the wrong version the log might show something like this:
INFO name.cache.amazonaws.com: 6380 failed to nominate (Faulted)
INFO > UnableToResolvePhysicalConnection on GET 33488
  • AccessKey missing in connection string: The access key might need to be put inside connectionString value. I have blogged about this issue before see here
  • Intermittent timeout issues: There might be intermittent timeout issues when Sitecore is connected to Redis. This KB article provides a good start to resolve these kind of issues. If this happens the log will show something like this:
Exception: System.TimeoutException
Message: Timeout performing EVAL, inst: 1, mgr: Inactive, err: never, queue: 24, qu: 0, qs: 24, qc: 0, wr: 0, wq: 0, in: 12544, ar: 0, IOCP: (Busy=5,Free=395,Min=200,Max=400), WORKER: (Busy=4,Free=396,Min=88,Max=400), clientName: client
Source: StackExchange.Redis.StrongName
   at StackExchange.Redis.ConnectionMultiplexer.ExecuteSyncImpl[T](Message message, ResultProcessor`1 processor, ServerEndPoint server)
   at StackExchange.Redis.RedisBase.ExecuteSync[T](Message message, ResultProcessor`1 processor, ServerEndPoint server)
   at StackExchange.Redis.RedisDatabase.ScriptEvaluate(String script, RedisKey[] keys, RedisValue[] values, CommandFlags flags)
   at Sitecore.SessionProvider.Redis.StackExchangeClientConnection.<>c__DisplayClass12_0.<Eval>b__0()
   at Sitecore.SessionProvider.Redis.StackExchangeClientConnection.RetryForScriptNotFound(Func`1 redisOperation)
   at Sitecore.SessionProvider.Redis.StackExchangeClientConnection.RetryLogic(Func`1 redisOperation)
   at Sitecore.SessionProvider.Redis.StackExchangeClientConnection.Eval(String script, String[] keyArgs, Object[] valueArgs)
   at Sitecore.SessionProvider.Redis.RedisConnectionWrapper.TryTakeWriteLockAndGetData(String sessionId, DateTime lockTime, Object& lockId, ISessionStateItemCollection& data, Int32& sessionTimeout)
   at Sitecore.SessionProvider.Redis.RedisSessionStateProvider.GetItemFromSessionStore(Boolean isWriteLockRequired, HttpContext context, String id, Boolean& locked, TimeSpan& lockAge, Object& lockId, SessionStateActions& actions)
  at Sitecore.SessionProvider.Redis.RedisSessionStateProvider.GetItemExclusive(HttpContext context, String id, Boolean& locked, TimeSpan& lockAge, Object& lockId, SessionStateActions& actions)
   at System.Web.SessionState.SessionStateModule.GetSessionStateItem()
   at System.Web.SessionState.SessionStateModule.BeginAcquireState(Object source, EventArgs e, AsyncCallback cb, Object extraData)
   at System.Web.HttpApplication.AsyncEventExecutionStep.System.Web.HttpApplication.IExecutionStep.Execute()
   at System.Web.HttpApplication.ExecuteStep(IExecutionStep step, Boolean& completedSynchronously)

Designing for performance

There are many factors which impact the performance of Redis. The only way to determine the best configuration for a certain site is to perform a load test with a load that is similar to production traffic. Based on my experience I recommend exploring below options:

  • Enable Clustering: It is often more effective to create a Redis cluster with multiple instances than to increase the size of a single non clustered Redis instance. Each Redis instance can only be scaled vertically by allocating more resources to it. With a cluster Redis will create multiple instances and divide the data over the instances based on its key. This technique is also referred to as sharding and is supported by Redis, which makes it transparent to Sitecore. Therefore there are no changes needed on Sitecore’s side, it just needs to have its Redis connection string pointed to the endpoint of the cluster.
    • Important note: Sitecore is using StackExchange.Redis.StrongName to access Redis. “Move” exceptions can occur below version 1.1.603 of this library when clustering is enabled. A little more information about this issue can be found here. This link only describes the issue in Azure but the same issue can occur anywhere else as well. Per below table all Sitecore 9.0 versions use a version of the Stackexchange Redis driver below 1.1.603 and might throw “Move” exceptions when configured to use a Redis cluster.
      Sitecore StackExchange Redis
      9.0 Initial Release (171002) 1.0.488
      9.0 Update-1 (171219) 1.0.488
      9.0 Update-2 (180604) 1.0.488
      9.1 Initial Release (001564) 1.2.6
      9.1 Update-1 (002459) 1.2.6
  • Keep compression disabled: the Redis server is single-threaded. This makes it perform well with small key-value pairs, but performance will decrease when the size of the data it stores goes up. The advantage of disabling compression is that Sitecore does not need to spend CPU time compressing and decompressing the data. However the amount of data that needs to be send to Redis goes up, we have seen the amount of data send to Redis triple without compression. This had a significant adverse impact on Redis’ performance and the performance of the entire site. The extra CPU time with compression enabled was negligible compared to overall CPU. Below image taken from Redis.io shows how throughput decreases with increased data size.

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out /  Change )

Google photo

You are commenting using your Google account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s