Skip to content

Don't run multiple "all" nodes in a cluster  #7074

@jstirnaman

Description

@jstirnaman

Is it valid to run multiple cluster members in "all" mode?
No, it is not valid.

I'm not sure the software would detect and reject if you attempted it. but running multiple compactors will almost certainly cause issues.

The documentation is contradictory

https://docs.influxdata.com/influxdb3/enterprise/admin/clustering/#cluster-architecture-examples :
# Nodes 9-10: Dedicated compactors
https://docs.influxdata.com/influxdb3/enterprise/get-started/multi-server/#high-availability-cluster :
Only one node can be designated as the Compactor. Compacted data is meant for a single writer, and many readers.

Slack Message

Recommendations

Run only one compactor node per cluster

clustering.md showed architecture examples with multiple nodes in compact or all mode, contradicting multi-server.md's correct statement that only one compactor node is allowed per cluster. Running multiple compactors causes data corruption.

Don't assign an all node in a multi-node cluster

Use --mode=all for a single-node Enterprise cluster. Avoid using an all node in a multi-node cluster. Some cluster features such as replication and catalog refresh aren't designed to work with single-node, all mode, clusters.

Compactor node must use the same node-id as the all node being replaced

Observe the following for all examples scaling from all mode to a multi-node cluster:

Replace the all node with nodes that have explicit, specialized modes
Assign compact mode to exactly one node that uses the same node-id as the all node being replaced

Use real code or configuration in code blocks

Fix the notation/syntax in this "bash" example so that it represents real code.

Metadata

Metadata

Labels

product:v3-monolithInfluxDB 3 Core and Enterprise (single-node / clusterable)

Type

Projects

No projects

Milestone

No milestone

Relationships

None yet

Development

No branches or pull requests

Issue actions