How to remove/shrink storage node quantity in vm cluster? about victoriametrics HOT 2 CLOSED

Hello @naughtyGitCat

There are a few ways to achieve this:

1. Separate read and write paths to use different sets of storage nodes.
   You can change vminsert configuration to write data to 3 nodes out of 6 existing nodes. vmselects should still read the data from all 6 nodes. Once retention period removes all data from excessive storage nodes it's possible to remove those storage nodes and update vmselect configuration.
   This option works well for short retention periods as it requires all storage nodes to be available until retention will remove the data.

2. Re-import data to be distributed across 3 storage nodes.
   The idea is to update vminsert configuration to only write data to 3 storage nodes which will be left in cluster.
   After this you can export data from cluster and import back.
   Please, make sure that deduplication is enabled before importing data. You can use dedup.minScrapeIntervals=1ms for this case.
   In this case data will be distributed across 3 storage nodes used in vminsert configuration, deduplication will remove the duplicate samples created by import operations in the background.
   After import is finished it is possible to remove unneeded storage nodes as the data for these storage nodes is already present in cluster.

3. Create new cluster with the desired configuration and migrate the data.
   This option is the most straightforward and simple to implement. It works unless it is required to update an existing cluster.

from victoriametrics.

thanks, your suggestion is very meticulous

from victoriametrics.