Built a distributed counter that survives network partitions using node-local counters and eventual consistency. No fancy CRDTs needed - just good old divide and conquer.
Each node owns its chunk of the counter
Adds are super fast (just local increment)
Reads aggregate from all nodes concurrently
Network partition? No problem, nodes keep counting
localValue with mutex for thread safety
KV store for persistence
Concurrent reads using goroutines
Simple message types: add, read, local
Partition Friendly: Nodes keep working even when isolated
Eventually Consistent: Global reads converge once network heals
Fast Writes: No coordination needed for increments
Clean Code: ~150 lines tells the whole story
Pros:
Simple to reason about
Naturally partition tolerant
Local writes are instant
No complex state merging
Cons:
Reads hit all nodes
Network issues = incomplete reads
Not strictly consistent (but that's fine)
Could've gone full CRDT, but letting each node count its own stuff just works. Plus, Go's goroutines make those concurrent reads feel like cheating.
