Lab: Store Gateway, Compactor and downsampling

What you will learn in this TP

• Serve the history
• Configure retention
• Inspect the bucket
• Query long periods

History and compaction #

What you will learn in this section

• Serve the history
• Configure retention
• Inspect the bucket
• Query long periods

Context. Recent data is served by the sidecars, but the history lives in object storage. You add the Store Gateway to read it and the Compactor to compact, downsample and apply retention.

Prerequisites: previous labs  ·  Estimated time: 45 min

Store and Compactor.

1. 1. Start the Store Gateway

   Why? Without a Store Gateway, the archived blocks exist but nothing can read them: queries cannot go back further than local retention.

   Add it to Query with --endpoint=store:10901.

2. 2. Start the Compactor

   Why? The compactor reduces the number of blocks and creates lower resolutions so you can keep years of history at low cost; only one instance must write to the bucket.

   Retention: 30 days raw, 6 months at 5 min, 2 years at 1 h. Why only one instance?

3. 3. Inspect the bucket

   Why? Inspecting the bucket shows the actual blocks (time range, resolution) and helps with diagnosis when a period seems to be missing.

   Display the list of blocks with their resolution.

4. 4. Query the history

   Why? Automatic downsampling speeds up queries over long periods by reading fewer data points.

   Run a 7-day query with automatic resolution.

Difficulty level : ●●●○○ (3/5)