// the find
ContentSquare/chproxy
Open-Source ClickHouse http proxy and load balancer
chproxy sits in front of ClickHouse and handles load balancing, query queueing, caching, and user/network access control via a YAML config. It's aimed at teams running ClickHouse at scale who need to expose it to multiple users or services without directly opening ClickHouse ports. ContentSquare built it internally and open-sourced it.
The query result cache is a real implementation — filesystem or Redis backends, async deduplication so concurrent identical queries hit ClickHouse only once. Per-user and per-cluster-user concurrency limits with queuing are genuinely useful when you have dashboards hammering the same instance. Config hot-reload and Prometheus metrics are both there out of the box. The test coverage on the cache layer in particular looks serious, with transaction registry tests across both in-memory and Redis backends.
The README is essentially just a link to an external docs site, which is a bad sign for longevity — if that domain lapses, the project becomes opaque. The project originated from one company's internal needs and the commit cadence reflects it; forks are high but contributor diversity is narrow. No native support for ClickHouse Keeper or the newer replicated table routing that matters if you're on a modern cluster setup. Redis dependency for distributed caching is unavoidable if you run multiple proxy instances, but there's no guidance on what happens when Redis goes down — whether it silently falls back to pass-through or starts refusing requests.