Will TTL play a role in cleaning up old values if replace before TTL?

[ShanmugamC]

Hi Team,
If we set higher TTL for a key and end up replacing value for it before TTL is over multiple times, what happens to old values as i see some memory leak for this scenario?

E.g:
Key 1: TTL : 12 hours but i replace the value 100 time within 12 hours, what happens in this case? Will it hold reference for old values still?

[ben-manes]

This depends on your expiration setting.

• expireAfterWrite will reset the entry's lifetime on every update, causing it to have a 12 hour duration.
• expireAfterAccess will reset the entry's lifetime on every read or write.
• expireAfter(expiry) will reset to the return value of Expiry.expireAfterUpdate.

If you want an entry to expire only after creation, regardless of subsequent reads or writes, then you should use expireAfter(expiry).

Caffeine.newBuilder()
    .expireAfter(new Expiry<Key, Graph>() {
      public long expireAfterCreate(Key key, Graph graph, long currentTime) {
        return TimeUnit.HOURS.toNanos(12);
      }
      public long expireAfterUpdate(Key key, Graph graph, 
          long currentTime, long currentDuration) {
        return currentDuration;
      }
      public long expireAfterRead(Key key, Graph graph,
          long currentTime, long currentDuration) {
        return currentDuration;
      }
    }).build();

[ShanmugamC]

sure , Thanks for your detail reply.

I noticed memory leak during this process as i mentioned in the question earlier, so my question is if i override the same key , will it still hold reference with older value and cause memory leak?

[ben-manes]

Oh, sorry. I misread that then.

The old value is removed. If you have a listener configured then it will be notified. The cache is merely a fancy ConcurrentHashMap so it should follow those expectations.

[ShanmugamC]

Thanks for your reply @ben-manes !

[ShanmugamC]

I am Initiating cache as below:

// Class CacheRegion holds below

       AsyncCache<K, V> asyncCache = Caffeine.newBuilder()
                .maximumSize(5000)
                .expireAfterWrite(Duration.ofHours(12))
                .recordStats()
                .buildAsync();

//And i am overriding same cache multiple times before cache expiring using below code with spring webflux

  CompletableFuture<CacheObject<T>> cf = CompletableFuture.completedFuture(cacheObject);
  asyncCache.put(key, cf);

// I was able to reproduce CacheRegion holding lot of memory for every-time i override the same key. With 12 hours TTL slowly memory builds up and causing OOM , but to confirm this i reduced TTL to 30 mins and after 30 mins JVM memory usage is stable and i don't see the issue.

Just wanted to understand if i am doing something wrong here and why old keys/values still stay after overriding with new key & value.

[ben-manes]

That looks correct. The cache internally is a Map<K, CompletableFuture<V>> and it won't hold onto the old value.

fyi, you can use asyncCache.synchronous.put(key, cacheObject) which wraps the value for you.