HomeGuidesIceberg

Apache Iceberg Table Maintenance Checklist

Shachar Shamir
October 8, 2026
6
min read

Apache Iceberg keeps a full history of table changes. That history enables time travel and safe concurrent writes, but it also accumulates files that need regular cleanup.

Expire old snapshots

Each commit creates a snapshot. Expiring snapshots older than your retention window removes references to data files that are no longer needed.

Remove orphan files

Failed or interrupted writes can leave files that no snapshot references. A periodic orphan-file cleanup reclaims that storage.

Compact data files

Schedule compaction

Run compaction on partitions that receive frequent writes, starting with the bin-pack strategy.

Target file size

Keep a consistent target file size, typically between 128 MB and 512 MB, depending on your query engine.

Rewrite manifests

Many small manifest files slow down query planning. Rewriting manifests groups them so planning stays fast as the table grows.

[[cta:contact]]

Key takeaways

Automate snapshot expiry, orphan cleanup, compaction and manifest rewrites so table performance and storage costs stay predictable.

‍

On this page
See DualBird accelerate your own Spark workloads.
Book a demo

Related guides

Iceberg

Apache Iceberg Table Maintenance Checklist

The recurring maintenance tasks that keep Apache Iceberg tables fast and storage costs under control.

October 8, 2026
Iceberg

Iceberg Compaction Explained: Bin-Pack vs. Sort

Why Apache Iceberg tables need compaction, how bin-pack and sort strategies differ, and how to choose between them.

October 18, 2026
See 10x-30x faster Spark on your own workload
DualBird accelerates existing Spark and Iceberg jobs with no code changes.
Book a demo
How DualBird accelerates Spark and Iceberg
Purpose-built hardware, delivered like software, for big data processing on AWS.
Learn about the product →
Benchmark: Iceberg compaction, faster and cheaper
See how sort-based compaction ran faster at lower infrastructure cost, with no tuning.
Read the benchmark →
Questions about your pipeline?
Talk to an engineer about performance and cost for your workloads.
Contact us →

Transform your data infrastructure performance with a few clicks

Zero risk, zero effort, incredible results.