ClickHouse Rebalancing in OSS: Let’s Talk About the Gaps

ClickHouse offers powerful performance at scale — but if you're running it in an open-source, self-managed environment, one thing becomes clear once you hit a certain scale: there’s no easy way to rebalance shards.In this talk, we’ll share our experience rebalancing a large ClickHouse OSS cluster. We’ll walk through:

Why shard imbalance happens and why it matters What rebalancing support exists today — and what’s still missing in OSS How we designed a scriptable way to move parts across shards — tailored to our specific use case and not a one-size-fits-all solution Pitfalls we hit around ongoing merges, pending mutations, and active inserts — all of which made it difficult to ensure consistency and correctness during part movement. This is an engineer’s perspective — code-first, painfully real, and focused on what it takes to make ClickHouse scale cleanly when you're not using a managed service. We’ll close with thoughts on how the community could evolve rebalancing support in open-source deployments.

If you’ve ever dealt with skewed data, overloaded shards, or manual repair jobs, you’ll feel right at home.