How LakeOps Works
LakeOps is a control plane for Apache Iceberg. It connects to your existing catalogs, reads table metadata to decide what each table needs, and runs maintenance on your tables in place. This page covers the model behind the rest of the docs.
Three ways in, one backend
| Interface | Sign-in | Use it for |
|---|---|---|
| Console | Browser login | Connecting catalogs, reviewing health, setting policies. |
| REST API | Org API key | Scripts, CI, and SQL. |
| MCP server | Org API key | AI agents such as Cursor and Claude. |
Org admins create API keys in Organization > API Keys and choose their scopes. The same key works for the REST API and the MCP server. See scopes for what each one allows.
Data regions
LakeOps runs in data regions around the world and covers most major cloud regions. You choose a Data region when you log in. Each region keeps its own catalogs, so connect a catalog in the region you will manage it from and use that region every time.
The REST API and MCP server run in the same region, on your region's API host. The Connect Your Agent setup shows the exact URL for your region.
What LakeOps does in your storage
- •Discovery and analysis. When you connect a catalog, LakeOps scans its namespaces and tables. From then on it reads Iceberg metadata (snapshots, manifests, file statistics) to score health, raise insights, and measure how fast each table accumulates small files, snapshots, and delete files.
- •Maintenance. Operations run on the Rust compaction engine, built on Apache DataFusion. A job reads only the files it needs to rewrite, writes the results to your bucket, and commits a new snapshot to your catalog with the credentials you provided.
- •Cleanup. Snapshot expiry and orphan file cleanup delete files from your bucket, so the credentials need delete access for them to run.
Your tables stay in your storage and your catalog. Nothing is migrated or copied into LakeOps. The permissions LakeOps needs, with a sample IAM policy, are in Connect a Catalog.
Tables on Amazon S3 and S3-compatible storage are supported today. Support for Google Cloud Storage and Azure (Blob Storage and ADLS) is in alpha.
How operations run
Every maintenance operation runs in one of three ways. Results show up in the table's Events tab and in Data > Events.
Automatic
Runs when table signals cross thresholds. This is how Adaptive Maintenance works by default.
Scheduled
Runs on a 5-field cron expression, such as 0 2 * * * for 2:00 every day.
Configuration policies are different: they set Iceberg table properties on the tables they cover instead of running a maintenance job, so they have no schedule.
How policies apply to a table
A policy holds the settings for one operation type, such as Expire Snapshots, and the tables it covers. There are two kinds:
- •Table policies, saved from a table's Optimization tab. They apply to that table only.
- •Governance policies, created in Manage > Policies. They cover every catalog, one catalog, or a set of namespaces and tables. Namespace and table names can be exact names or regex patterns, and you can exclude namespaces and tables.
Which policy wins, per operation type
- A table policy, if the table has one.
- Otherwise, the most specific governance policy that matches the table.
- Among equally specific matches, the one with the higher priority value.
The Optimization tab shows which policy is active and which others also match. Turning an inherited policy off for one table adds that table to the governance policy's exclusion list, which changes the shared policy. Full details are in Policies.
Finding a table in the console
Open a table from Data > Explore, which browses catalogs and namespaces, or from a row in Data > Tables, which lists every table with its health. A table page has these tabs:
| Info | Schema, partition specs, sort orders, location, format, and table properties. |
| Snapshots | Snapshot history with tag, branch, rollback, and set-current actions. |
| Metrics | Record and file trends over time, file size distribution, and table growth. |
| Optimization | Per-table maintenance settings, Execute, and inherited policies. |
| Maintenance | Adaptive signals: scores, accumulation rates, projected triggers. |
| Insights | Findings for this table, by severity. |
| Events | Every operation on this table, with its status and results. |
| Partitions | Per-partition file and size breakdown. |
| Query | Run SQL against the table. |
| Settings | Rename, view the CREATE statement, export metadata, delete the table. |
