<?xml version="1.0" encoding="UTF-8"?>
<rss version="2.0">
<channel>
  <title>Lakehouse</title>
  <link>https://lakehouse.website/</link>
  <description>Field notes on data engineering.</description>
  <language>en</language>
  <lastBuildDate>Fri, 09 Oct 2026 08:00:00 +0000</lastBuildDate>
  <item>
    <title>Exactly-once from Kafka into ClickHouse</title>
    <link>https://lakehouse.website/posts/kafka-clickhouse-exactly-once</link>
    <guid>https://lakehouse.website/posts/kafka-clickhouse-exactly-once</guid>
    <pubDate>Fri, 09 Oct 2026 08:00:00 +0000</pubDate>
    <description>Kafka transactions stop at the topic. Getting the same guarantee into ClickHouse means making inserts repeatable and letting the database throw away retries.</description>
  </item>
  <item>
    <title>Iceberg maintenance: four jobs, in order</title>
    <link>https://lakehouse.website/posts/iceberg-maintenance</link>
    <guid>https://lakehouse.website/posts/iceberg-maintenance</guid>
    <pubDate>Fri, 21 Aug 2026 08:00:00 +0000</pubDate>
    <description>Compaction gets the attention, but snapshot expiry and orphan cleanup are what keep the storage bill flat. The order we run them in, and the settings we actually touch.</description>
  </item>
  <item>
    <title>Late data in Flink: allowed lateness or a side output?</title>
    <link>https://lakehouse.website/posts/flink-late-data</link>
    <guid>https://lakehouse.website/posts/flink-late-data</guid>
    <pubDate>Sun, 14 Jun 2026 08:00:00 +0000</pubDate>
    <description>Both keep late events from vanishing. One keeps window state around and re-fires results; the other hands the problem to you. Choose by who reads the output.</description>
  </item>
  <item>
    <title>ClickHouse materialized views are insert triggers</title>
    <link>https://lakehouse.website/posts/clickhouse-materialized-views-are-triggers</link>
    <guid>https://lakehouse.website/posts/clickhouse-materialized-views-are-triggers</guid>
    <pubDate>Mon, 30 Mar 2026 08:00:00 +0000</pubDate>
    <description>They run on each inserted block, not on the table. Once that clicks, the surprises around backfills, joins and deletes stop being surprises.</description>
  </item>
  <item>
    <title>Partition for deletion, sort for reading</title>
    <link>https://lakehouse.website/posts/partition-for-deletion-sort-for-reading</link>
    <guid>https://lakehouse.website/posts/partition-for-deletion-sort-for-reading</guid>
    <pubDate>Wed, 12 Nov 2025 08:00:00 +0000</pubDate>
    <description>In ClickHouse and Iceberg alike, the partition key decides what you can drop cheaply and the sort key decides what you can read cheaply. Mixing them up is expensive.</description>
  </item>
</channel>
</rss>
