Observe on Apache Iceberg: Unlocking Open Observability

skip\_to\_content

Blog/ Product and Technology/Unlocking Open Observability with Observe on Apache Iceberg

Jul 30, 2026/6 min readProduct and Technology

Copy post link

Open in ClaudeOpen in ChatGPT

Unlocking Open Observability with Observe on Apache Iceberg


!Digital illustration of an iceberg in a ring with a blue background with white dots

We are excited to announce that Observe on Apache Iceberg™ is now in private preview. Your observability data can now be stored as Apache Iceberg tables in your own Amazon S3 bucket with the full Observe experience running on top. This is part of Observe by Snowflake’s broader commitment to open standards in observability. In addition to using OpenTelemetry-native ingestion, your data now lives in a format any Iceberg-compatible engine can read, in storage that you own and control.

The volume of logs, metrics and traces generated by modern systems has grown dramatically. Leading enterprises now ingest massive amounts of data daily. Yet traditionally that data has almost always lived in a silo — encoded in vendor-specific formats, queryable only through the vendor's interface and inaccessible to the rest of the organization. Engineering teams troubleshoot in one tool while business and data teams work in another, and joining telemetry signals with business data requires ETL, manual exports or a custom pipeline. When your needs outgrow the vendor’s platform, or when you want to run analytics on top of your telemetry, getting that data out is often harder than it should be.

These issues are a consequence of how traditional observability platforms were architected: built to serve a single vendor's interface, with no expectation that the data would ever need to live anywhere else. Open table formats such as Apache Iceberg change that. When your telemetry is stored in Iceberg, it lives alongside the rest of your data, under your control and accessible through the tools your teams already use. This minimizes the need to extract or replicate data into a separate system — and the related costs of that process.

Your data, in your object storage, in an open format

With Observe on Iceberg, your telemetry is written as Iceberg tables directly to an Amazon S3 bucket in your AWS account. Observe accesses that bucket through an IAM role you create and control. If you revoke that role or remove the data lake configuration, Observe loses access. The data stays in your bucket in an open format, yours to keep. You can configure multiple data lakes, and data sets across them are designed to remain isolated.

Observe on Iceberg is built on Snowflake-managed Iceberg tables using the v3 table specification. Because these tables are managed, compaction, file sizing, and snapshot and metadata upkeep all happen automatically — as a result, you get high-performance queries without having to maintain the tables yourself. The v3 spec also brings new table capabilities, including richer data types and performance improvements.

!Architecture diagram for Observe on Iceberg.

Figure 1: Architecture diagram for Observe on Iceberg.

In the private preview, you can:

• Ingest data into Iceberg via the Observe pipeline without a separate export or ETL step: Your existing ingestion setup is unchanged. Data flows through the Observe ingestion pipeline and is written into standard Iceberg tables in your own S3 bucket.

• Store observability data in an open format on your own object storage: Data sets are written as standard Iceberg tables in your S3 bucket, under your governance, retention policies and access controls. If you remove Observe's access, the data remains yours.

• Build monitors, dashboards and correlations on top of your data within Observe: The full Observe experience — monitoring, dashboards, OPAL transformations, AI SRE and correlations — is designed to work identically on Iceberg-backed data sets. Nothing changes about how you use Observe day to day.

!Video Thumbnail

Click for sound

0:53

Why Iceberg is the right format for observability data

Observability data is well-suited to columnar storage, and Iceberg stores data in Parquet by default. Time stamps, service names, severity levels and structured attributes repeat across millions of events in patterns that compress efficiently and scan well with partition pruning.

But Iceberg is more than just Parquet. Parquet provides columnar files that are easy to query if you already know where they are. Iceberg adds the table layer on top, providing schema evolution, partitioning, snapshots and a catalog that any compatible engine can discover and read without custom integration. That's what makes your telemetry open and accessible instead of just stored in an efficient format.

In our testing, observability workflows also run with similar performance on Iceberg tables as on Observe-native data, so you don’t need to choose between openness and performance. Engineering teams can continue using Observe for incident investigation, dashboards, monitors and AI-assisted root cause analysis, while the rest of your organization gets direct access to the same data, in the tools they already use.

Query with any engine

Once your data is in Iceberg tables, it isn't readable only through Observe. Observe exposes a read-only Iceberg REST Catalog on the Observe API so that any Iceberg-compatible query engine can discover and read your tables directly. This interoperability is...