What I learned from a month with WalShadow on ClickHouse Managed Postgres

A database engineer evaluated ClickHouse’s managed WalShadow replication during its 2026 private preview against ClickPipes for synchronizing ClickHouse Managed Postgres changes to ClickHouse Cloud. WalShadow, which reads physical PostgreSQL WAL directly and writes decoded changes to ClickHouse, maintained roughly 0.4-second median latency from 1,000 to 200,000 rows per second, while ClickPipes latency depended on its sync interval and increased sharply above an observed throughput of about 50,000 rows per second. Both methods ultimately produced destination data matching the source, but WalShadow also continued automatically through schema changes, source restarts and resizes, destination outages, and auto-idling in the tests. The evaluation found caveats involving temporary cross-table transaction visibility, handling of CDC row versions and deletes in materialized views, fixed metadata values for initial-load rows, dropped destination tables, and delays caused by large transactions. Issues discovered during testing, including type changes, reconnection after resizing, recovery after destination restarts, and replication of certain tables without primary keys, were reported and subsequently fixed. The author concludes that ClickPipes remains appropriate for externally hosted PostgreSQL sources, whereas WalShadow is particularly suited to near-real-time analytics for ClickHouse Managed Postgres, though additional operational monitoring and notifications would improve it.

Use this post, company, and trend context to find content marketing opportunities, perform competitive analysis, or address product feature gaps via the Plushcap MCP server or the Plushcap API.