ClickHouse's long-term support line settles down as dbt-core opens the door to Exasol

ClickHouse's LTS lines get a coordinated sweep

ClickHouse spent the week doing what mature analytics engines do best: paying down maintenance debt across several long-term support branches at once, from the newest LTS line down through the previous stable branch, its follow-on patches, the midweek fixes, an earlier maintenance cut, and the oldest supported line, plus the earliest patch of the run. The practical upshot for anyone running ClickHouse in production: no matter which LTS branch you're pinned to, there's a current patch available, and the fix cadence suggests the project is deliberately keeping every supported line within reach of the same bug fixes rather than forcing upgrades. If you're on a pinned LTS, this is the week to schedule a rolling patch — but expect the notes to be incremental rather than headline-grabbing.

Managed Postgres gets a real onboarding path

For teams who use ClickHouse alongside operational Postgres, the new getting started experience for ClickHouse Managed Postgres walks users through a four-step flow that goes from provisioning a cluster to querying operational data and running real-time analytics inside ClickHouse. It's a small thing on paper and a large one in practice: the hardest part of adopting a managed analytics service is usually the first hour, and a guided path from "I have a database" to "I have a dashboard" removes most of the reasons teams stall.

The week's other ClickHouse posts round out the ecosystem picture. The .NET Driver's road from 1.0 to 1.3 brought type-safe POCO workflows, extensible serialization, broader type coverage, and performance work, alongside official ecosystem integrations — meaning C# shops no longer have to hand-roll their mapping layer. More interesting for anyone who has ever been burned by a flaky test suite, ClickGap applies autonomous QA to merged ClickHouse changes: it executes reproducers, filters out false positives, and attributes regressions to specific commits. And on the broader Postgres front, read-your-writes semantics via WAIT FOR in PostgreSQL 19 lets individual reads block on a specific WAL position, which finally makes asynchronous replicas safe for the read-after-write patterns that analytics dashboards depend on.

Tinybird rebalances query budgets, Dagster tidies orchestration

Tinybird raised the default Query API and Pipe endpoint timeouts on paid plans, tiering them at ten seconds on free, twenty on shared paid infrastructure, and thirty on dedicated infrastructure, with existing custom overrides left untouched. The same update added a replica column to pipe_stats_rt, so you can see which ClickHouse replica executed each query — genuinely useful when you're debugging why the same pipe is fast one minute and slow the next. Over in Dagster, dynamic partitions can now be wiped and deleted in one action, which also clears the asset's degraded health status, and alert policies can now fire on successful code location deploys rather than only failures.

dbt-core's 2.0 preview widens the adapter map

dbt-core's 2.0 development line landed a Phase 1 Exasol adapter with a full macro bundle — connection and auth with TLS and timeout profiles, agate and seed type conversion including HASHTYPE, typed SQL literals, reserved-keyword auto-quoting, and materializations spanning table, view, incremental append/merge/delete+insert/microbatch, plus snapshots and model contracts. The same work introduced parse-safe views like dbt.models and dbt.columns, so querying an exported index can no longer silently surface a compile-only or never-populated column — a subtle correctness fix that will save warehouse teams a class of confusing empty results. It also hardens tarball extraction against symlink and hard-link attacks, includes invocation IDs in Databricks query comments, and fixes a log-format-file regression.