{"collection":{"slug":"data-engineering","name":"Data & Analytics Engineering"},"digest":{"id":"cwd_spYhtcId01TH1f2NQ85kr","weekStart":"2026-08-03","title":"ClickHouse flood and a Dagster retry-status signal","intro":"ClickHouse shipped its 26.7 release with major query and JOIN speedups, plus four stable-branch patches and deep dives into autoscaling and Managed Postgres resilience. Dagster clarified retry semantics with a new asset health status and a quarantine-preservation improvement from Tinybird.","body":"### The ClickHouse 26.7 story\n\nThe week belonged to [ClickHouse's 26.7 release](/release/rel_Qiekb4-WEdF24z5IN689K-clickhouse-release-26-7), which bundles tangible wins for query-heavy workloads. The headline is a speedup for `GROUP BY ... ORDER BY ... LIMIT` patterns — the kind of query that powers dashboards and top-N reporting — alongside three JOIN improvements and four vector search enhancements. Position-aware phrase search is a welcome addition for full-text work, and `EXPLAIN ANALYZE` finally gives you a single command to see both the plan and actual execution stats. Unified URL access rounds out the release by making it simpler to point ClickHouse at remote files over different protocols.\n\nThe company also published two engineering writeups that explain the \"why\" behind recent changes. The first details how [ClickHouse Cloud rebuilt its autoscaling](/release/rel_5FB_Je0_nmMfRlogEYbfE-fixed-cadence-to-seconds-making-clickhouse-cloud-autoscaling-more-reactive) on Kubernetes' controller-runtime with a ClickHouse-powered signals table, adding a reactive fast path that scales services up in seconds rather than waiting for the next scheduled pass — a meaningful improvement for spiky workloads. The second is a [clear explainer on WAL backpressure](/release/rel_lfB7rzqZxdyLM_uIsa16f-what-is-wal-backpressure-and-why-does-clickhouse-managed-postgres-need-it) in Managed Postgres: how slowing client writes protects disk space and gives the archiver room to recover. It's worth reading even if you're not a Managed Postgres user, because the pattern applies to any system with a write-ahead log.\n\n### Managed Postgres polish\n\nSeparately, [ClickHouse Managed Postgres got a feature batch](/release/rel_7Z3qaM6eWZ0L7yRDnyjMN-what-s-new-in-clickhouse-managed-postgres-customer-notifications-better): proactive customer notifications, richer observability, faster backups, and expanded extension support. None of these are flashy, but they collectively lower the operational burden for teams running Postgres alongside ClickHouse — especially the backup speedup, which shrinks the window where you're most exposed.\n\n### Patch cadence, as usual\n\nClickHouse issued five stable-branch patches this week: v26.7.2.59-stable, v26.6.2.160-stable, v26.5.6.113-stable, v26.3.17.110-lts, and v25.8.29.51-lts, plus an earlier v26.6.2.158-stable. These are routine maintenance releases covering bug fixes across the supported lines; no single change stands out, but staying current on your branch is still the right move.\n\n### Dagster's retry clarity and Tinybird's quarantine fix\n\nAway from ClickHouse, [Dagster 1.13.17](/release/rel_Oi5NYcjGitJS7QkzwspmR-1-13-17-core-0-29-17-libraries) made a small but meaningful change to asset health reporting: an asset whose materialization failures are all pending an automatic retry now shows a `WARNING` status instead of `DEGRADED`. The corresponding Slack, Teams, and email alerts now render a \"Failed, retry pending\" line, so you can tell at a glance whether an asset is actively recovering or genuinely broken. The UI also got a better asset graph sidebar search — it now matches only what's shown, searches the full asset key, sorts alphabetically, and disambiguates names with namespaces.\n\nTinybird's changes are quieter but practical: [preserving quarantine data across compatible deployments](/release/rel_H7sOR2tuH8BUcne0af0wq-persistent-quarantine-data-and-workspace-usage-trends) means you no longer lose quarantined rows when a Data Source changes its sorting key or column types (as long as the column set is unchanged), saving a support ticket to recover them. And the Organization Workspaces table now shows seven-day sparklines for vCPU time, memory, and errors, giving you a quick trend read without digging into individual nodes.\n\n### The week's shape\n\nThe dominant theme is ClickHouse doubling down on performance engineering — both in the 26.7 feature set and in the infrastructure behind its managed offerings. Dagster and Tinybird both made smaller, sharper improvements to feedback loops: Dagster telling you a failure is pending retry rather than degraded, Tinybird keeping your quarantine data intact. All of it points in the same direction: less guessing, more signal.","releaseIds":["rel_Qiekb4-WEdF24z5IN689K","rel_5FB_Je0_nmMfRlogEYbfE","rel_lfB7rzqZxdyLM_uIsa16f","rel_7Z3qaM6eWZ0L7yRDnyjMN","rel_Oi5NYcjGitJS7QkzwspmR","rel_H7sOR2tuH8BUcne0af0wq"],"releaseCount":13,"generatedAt":"2026-08-10T06:20:22.722Z","releases":[{"id":"rel_Qiekb4-WEdF24z5IN689K","title":"ClickHouse Release 26.7","path":"/release/rel_Qiekb4-WEdF24z5IN689K-clickhouse-release-26-7","org":{"slug":"clickhouse","name":"ClickHouse"},"importance":null},{"id":"rel_5FB_Je0_nmMfRlogEYbfE","title":"Fixed cadence to seconds: making ClickHouse Cloud autoscaling more reactive","path":"/release/rel_5FB_Je0_nmMfRlogEYbfE-fixed-cadence-to-seconds-making-clickhouse-cloud-autoscaling-more-reactive","org":{"slug":"clickhouse","name":"ClickHouse"},"importance":null},{"id":"rel_lfB7rzqZxdyLM_uIsa16f","title":"What is WAL backpressure, and why does ClickHouse Managed Postgres need it?","path":"/release/rel_lfB7rzqZxdyLM_uIsa16f-what-is-wal-backpressure-and-why-does-clickhouse-managed-postgres-need-it","org":{"slug":"clickhouse","name":"ClickHouse"},"importance":null},{"id":"rel_7Z3qaM6eWZ0L7yRDnyjMN","title":"What's new in ClickHouse Managed Postgres: Customer notifications, better observability, faster backups, extensions, and more","path":"/release/rel_7Z3qaM6eWZ0L7yRDnyjMN-what-s-new-in-clickhouse-managed-postgres-customer-notifications-better","org":{"slug":"clickhouse","name":"ClickHouse"},"importance":null},{"id":"rel_Oi5NYcjGitJS7QkzwspmR","title":"1.13.17 (core) / 0.29.17 (libraries)","path":"/release/rel_Oi5NYcjGitJS7QkzwspmR-1-13-17-core-0-29-17-libraries","org":{"slug":"dagster","name":"Dagster"},"importance":null},{"id":"rel_H7sOR2tuH8BUcne0af0wq","title":"Persistent quarantine data and workspace usage trends","path":"/release/rel_H7sOR2tuH8BUcne0af0wq-persistent-quarantine-data-and-workspace-usage-trends","org":{"slug":"tinybird","name":"Tinybird"},"importance":null}]}}