Does FusionCache leverage dedicated database in Redis

[jundayin]

Hello,

A weird behavior I encountered when using FusionCache: In my application I setup Memory cache and the Distributed cache which is targeting one Redis cluster in Azure. I call Fusioncache GetOrSet() to retrieve the cache entry with a key. The key is available in our Redis instance with ttl -1 but the Fusion logs just says distributed entry not found and had to fallback to factory to retrieve the data from DB everytime. Not sure if I misunderstand the functionality correctly, but my original expectation is when Fusion Cache reads L2 Redis it will retrieve the existing key from Redis directly.

Some small repros

business code

await _fusionCache.GetOrSetAsync<T>(
  CacheUtilities.BuildCacheKey(orgName, tenantName, serviceType),
  async _ =>
  {
     // await calling DB
  },
  options =>
  {
    if (conditional check if we don't do caching at all)
    {
      options.SetSkipDistributedCacheRead(true);
      options.SetSkipDistributedCacheWrite(true, true);
      options.SetSkipMemoryCacheRead();
      options.SetSkipMemoryCacheWrite();
    }
});

Startup

RedisCacheSettings redisCacheSettings = Configuration.GetSection(nameof(RedisCacheSettings)).Get<RedisCacheSettings>();
services
  .AddFusionCache()
  .WithDistributedCache(new Microsoft.Extensions.Caching.StackExchangeRedis.RedisCache(new RedisCacheOptions() { Configuration = redisCacheSettings.ConnectionString }))
  .WithSerializer(new FusionCacheSystemTextJsonSerializer())
  .WithBackplane(new RedisBackplane(new RedisBackplaneOptions { Configuration = redisCacheSettings.ConnectionString }));

And the fusion cache trace log says

info: ZiggyCreatures.Caching.Fusion.FusionCache[0]
      FUSION [N=FusionCache I=0HNCHJ48LQ21Q] (O=0HNCHJ48LQ21T K=xxxxx): GetOrSetAsync<T> call FEO[DUR=30s LKTO=/ SKMR=N SKMW=N SKDR=N SKDW=N SKDRWS=N DDUR=/ JIT=0 PR=N SZ=/ FS=N FSMAX=1.00:00:00 DFSMAX=/ FSTHR=30s FSTO=/ FHTO=/ TOFC=Y DSTO=/ DHTO=/ ABDO=N SBN=N ABBO=Y AC=N]
dbug: ZiggyCreatures.Caching.Fusion.FusionCache[0]
      FUSION [N=FusionCache I=0HNCHJ48LQ21Q] (O=0HNCHJ48LQ21T K=xxxxx): [MC] memory entry not found
dbug: ZiggyCreatures.Caching.Fusion.FusionCache[0]
      FUSION [N=FusionCache I=0HNCHJ48LQ21Q] (O=0HNCHJ48LQ21T K=xxxxx): [MC] memory entry not found
dbug: ZiggyCreatures.Caching.Fusion.FusionCache[0]
      FUSION [N=FusionCache I=0HNCHJ48LQ21Q] (O=0HNCHJ48LQ21T K=xxxxx): [DC] distributed entry not found
dbug: ZiggyCreatures.Caching.Fusion.FusionCache[0]
      FUSION [N=FusionCache I=0HNCHJ48LQ21Q] (O=0HNCHJ48LQ21T K=xxxxx): calling the factory (timeout=/)
dbug: ZiggyCreatures.Caching.Fusion.FusionCache[0]
      FUSION [N=FusionCache I=0HNCHJ48LQ21Q] (O=0HNCHJ48LQ21T K=xxxxx): [MC] saving memory entry
dbug: ZiggyCreatures.Caching.Fusion.FusionCache[0]
      FUSION [N=FusionCache I=0HNCHJ48LQ21Q] (O=0HNCHJ48LQ21T K=xxxxx): [DC] saving distributed entry

[jundayin]

Currently no founding so far. I've checked the Redis instance and there is only one db (db0) in place

[jundayin]

Okay I create a brand new empty Azure Cached For Redis and rerun everything.

Apparently the cache key set by FusionCache contains "v2" prefix like v2:somecachekey which I didn't see it is mentioned in the documentation.

[jundayin]

Okay, after reading the docuementation again and read the source code a bit. I finally understand the v2 is actually the wire format of the cache key. It is by design

[jodydonetti]

Hi @jundayin orry for the delay, pretty busy these days.

Yes you are right, there's an additional "part" added to the cache key to handle versioning/evolution.

Based on your initial question, am I right that you tried to set the value in the cache NOT from another FusionCache instante?
I'm asking both to better understand your scenario, and because I'd like to point out something else: on top of having an extra "part" in the cache key, the value itself in the cache is also NOT the plain value you'd like to put there, because to support all the features of FusionCache there's the need for an "envelope" that contains additional metadata.

Think of it like this (simplified code):

class DistributedEntry<TValue> {
  int Metadata1 { get; set; }
  int Metadata2 { get; set; }
  int Metadata3 { get; set; }
  // ...
  TValue Value  { get; set; } // YOUR VALUE IS HERE
}

This means that, for example, if you save a Person in the distributed cache with FusionCache, and then try to read it NOT via FusionCache but directly, you can't expect to just deserialize the value in Redis into a Person.

Hope this helps, let me know.

[jundayin]

Hello @jodydonetti No worries. You have your main job to do

Anyway

Based on your initial question, am I right that you tried to set the value in the cache NOT from another FusionCache instante

You are right, the original value is set by StackExchange.Redis directly. We currently leverage the raw Redis operation

on top of having an extra "part" in the cache key, the value itself in the cache is also NOT the plain value you'd like to put there

I actually found that out myself after I used an empty testing Redis so I can very easily locate the cache value generated by FusionCache

Thank you

[jodydonetti]

You're welcome! Out of curiosity, now that we clarified this, does it make sense? Does it brake your plans on using FusionCache?
I'm interested in how people get to know it and if details like this change their plans.

Also, I've added to my backlog an item to update the docs to be more clear about this part of design, thanks for the input!

[jodydonetti]

Hi @jundayin , I just updated the docs to better explain the wire format envelope (the data structure containing the value + metadata) and the versioning identifier (that changes the cache key).

You can read here about the envelope and right below that, the updated part about versioning.

For both I included some practical examples to make it clearer.

I tried my best to be clear and exhaustive, but if you have some time to read it let me know what you think.

Thanks!