fix: add missing continue to shard cache put

[Longarithm]

If we are unable to remove element with hash taken from deletions queue, we remove LRU item from cache. This is not optimal - we should continue iterating over deletions queue until we don't exhaust it.

Testing

Existing tests - but we will have to add more #7498

[matklad]

Do we have some existing tests which cover this scenario? Can we extend them to include assertions about the state of the cache?

[matklad]

Good catch! If we already caught a bug here, let's try to compress this experssion to make it simpler and less error prone. Something like this should work I think:

            // First, try to evict value using the key from deletions queue.
            if let Some(key) = self.deletions.pop() {
                match self.cache.pop(&key) {
                    Some(value) => {
                        self.total_size -= value.len() as u64;
                        self.metrics.shard_cache_pop_hits.inc();
                    }
                    None => self.metrics.shard_cache_pop_misses.inc(),
                }
                continue;
            }

[mm-near]

+1 for adding the test (as we shouldn't be in a hurry now).

[mm-near]

please add the test.

[akhi3030]

I suspect this PR is no longer being actively worked on. Feel free to reopen when you are ready to start working on it again.