We had an incident where a Redis connection blip caused 200 concurrent requests to bypass cache and hammer PostgreSQL directly. The fix was surprisingly simple – one import from <a href="http://golang.org/x/sync/singleflight" target="_blank" rel="noopener nofollow">golang.org/x/sync/singleflight</a>.
I put together a fully runnable demo with:
<ul>
<li>Go service with cache-aside + singleflight pattern</li>
<li>Docker Compose stack (PostgreSQL, Redis, Prometheus, Grafana)</li>
<li>Pre-built Grafana dashboard so you can see the deduplication in real time</li>
<li>Load test script (200 concurrent requests x 25 waves)</li>
<li>Results: ~97% cache hit rate, only ~220 DB queries out of ~7,500 requests.
GitHub: <a href="https://github.com/augustus281/go-latency" target="_blank" rel="noopener nofollow">https://github.com/augustus281/go-latency</a></li>
</ul>

Augustus Nguyen

A Redis connection blip caused 200 concurrent requests to bypass cache and hit PostgreSQL directly. The solution was the singleflight pattern from golang.org/x/sync/singleflight, which deduplicates concurrent identical requests so only one hits the database. A runnable demo is provided with a full Docker Compose stack including PostgreSQL, Redis, Prometheus, and Grafana, plus a load test script. Results show ~97% cache hit rate with only ~220 DB queries out of ~7,500 total requests.

I built a demo showing how singleflight reduced our database load by 97% during a cache stampede

Really like how you showed the impact with actual metrics and a full setup, it makes it so much easier to appreciate what’s going on. That 97% hit rate and the huge drop in DB queries say it all. Definitely saving this one 🙌

if we have multiple instance would it work?