Set up Mimir on Grafana Labs

Grafana Mimir hardware requirements

Wed, 03 Jun 2026 09:01:40 +0200

Grafana Mimir hardware requirements

This page describes the hardware, storage, and infrastructure requirements for running Grafana Mimir in production. For component-level CPU, memory, and disk sizing, refer to Planning Grafana Mimir capacity.

Compute

The actual CPU and memory requirements depend on the number of active series, the rate of samples ingested per second, and query load. Refer to Planning Grafana Mimir capacity for per-component sizing guidance.

As a general guideline, provision nodes with a CPU-to-memory ratio of at least 1:4, meaning that for every CPU core, there should be 4GB of memory. For example, a node with 8 CPU cores should have at least 32 GB of memory. Ingesters and store-gateways are the most memory-intensive components.

Disk storage

Components such as ingesters, store-gateways, and compactors require fast and persistent, local disk resources to be available to the host machine storage for the write-ahead log (WAL), TSDB blocks, and temporary data during compaction.

Disk requirements

Grafana Mimir requires block storage devices that provide:

Low latency: SSD-backed storage is strongly recommended. Ingester write-ahead log (WAL) writes and store-gateway index lookups are latency-sensitive.
Adequate IOPS: A minimum of 50 IOPS per GB is recommended for ingester disks. Compactors also benefit from high IOPS during block compaction.
POSIX filesystem semantics: The filesystem must support standard POSIX operations including mmap, fsync, and file locking. Grafana Mimir’s TSDB engine relies on memory-mapped files for reading on-disk data.

Cloud provider examples:

Cloud provider	Recommended disk types
AWS	EBS `io1` or `gp3` SSD volumes
GCP	`pd-ssd` persistent disks
Azure	Premium SSD managed disks (`managed-csi-premium` storage class)

Unsupported filesystems

Warning
NFS and network filesystems are not supported for any Grafana Mimir component that uses local disk storage.

The following types of filesystems are not supported:

NFS (including Azure NFS file shares)
Amazon EFS
FUSE-based network filesystems
Any filesystem that does not provide full POSIX mmap consistency guarantees

Network filesystems do not provide the consistency guarantees that Grafana Mimir requires for memory-mapped I/O. Using them can result in:

Component crashes.
Silent data corruption.

For more background, refer to the Prometheus storage documentation.

Network

Grafana Mimir components communicate over gRPC and HTTP. For production deployments with high ingest or query throughput, a network bandwidth of 10 gigabit per second or faster between nodes is recommended.

For smaller or development deployments, lower bandwidth is acceptable, but network latency between components should remain low (ideally sub-millisecond within the same datacenter or availability zone).

Object storage

Grafana Mimir requires object storage for long-term block storage, ruler configuration, and alertmanager state. The following object storage services are supported:

Provider	Supported service
AWS	Amazon S3
GCP	Google Cloud Storage (GCS)
Azure	Azure Blob Storage
Self-hosted	Any S3-compatible API

Migrate to Grafana Mimir

Wed, 03 Jun 2026 09:01:40 +0200

Migrate to Grafana Mimir

Refer to the following migration guides when you migrate to Grafana Mimir.

Deploy Mimir with Helm

Wed, 03 Jun 2026 09:01:40 +0200

Deploy Mimir with Helm

The mimir-distributed Helm chart for Grafana Mimir and Grafana Enterprise Metrics allows you to configure, install, and upgrade Grafana Mimir or Grafana Enterprise Metrics within a Kubernetes cluster.

Use ingest storage

Grafana Mimir uses the ingest storage architecture by default when deployed with the Helm chart. You can configure an external Kafka backend or disable ingest storage if needed.

For details, refer to Run Grafana Mimir in production using the Helm chart, which provides complete setup and configuration guidance for production environments.

To learn more about the ingest storage architecture and related configuration options, refer to:

Example values files

The mimir-distributed Helm chart includes the following example values files:

File name	Description
`values.yaml`	Contains default values for testing GEM in non-production environments using a test MinIO deployment for object storage.
`small.yaml`	Contains values for a higher scale than defaults, for ingestion up to approximately one million series. Not suitable for high-availability production use, due to single replicas of key components.
`large.yaml`	Contains values for production use for ingestion up to approximately ten million series.

Deploy Grafana Mimir with Puppet

Wed, 03 Jun 2026 09:01:40 +0200

Deploy Grafana Mimir with Puppet

A puppet module is community maintained to deploy Grafana Mimir. This module use the deb or rpm packages to install Grafana Mimir.

Start using this module

Add this module to your Puppetfile:

mod 'ovhcloud-mimir', '1.0.0'

Install and configure Mimir

Call the main class of the module inside a puppet manifest like this:

class{'mimir': }

If called like this, this module will use the default configuration parameters of Mimir and therefore start mimir in standalone mode.

To customize the configuration, just add a hash with the appropriate configuration key like this:

class {'mimir':
    config_hash => {
        target => 'querier',
    },
}

Configuration parameters

# Mimir version under the form X.X.X or one of the supported puppet values
# as 'present', 'latest', ...
[package_ensure: <string> | default = "present"]

# Boolean to specify if module should manage mimir user
[manage_user: <boolean> | default = false]

# Home directory for the managed user
# Only if manage_user is set to true
[user_home: <string> | default = "/var/lib/mimir"]

# Binary to use as shell for managed user
# Only if manage_user is set to true
[user_shell: <string> | default = "/sbin/nologin"]

# Additionnal groups the managed user should be connected to
# Only if manage_user is set to true
[user_extra_groups: <array> | default =  []]

# Directory to store the mimir configuration
[config_dir: <string> | default = "/etc/mimir"]

# Group to use for configuration resources
[config_group: <string> | default = "mimir"]

# Hash containing the configuration keys to override
[config_hash: <hash> | default = {}]

# Owner to use for configuration resources
[config_owner: <string> | default = "mimir"]

# Additional arguments to set to the mimir process
[custom_args: <array> | default = []]

# Directory to store mimir logs if log to file is enabled
# Only if log_to_file is set to true
[log_dir_path: <string> | default = "/var/log/mimir"]

# Mode of the directory used to store logs
# Only if log_to_file is set to true
[log_dir_mode: <string> | default = "0700"]

# Filename to store mimir logs if log to file is enabled
# Only if log_to_file is set to true
[log_file_path: <string> | default = "mimir.log"]

# Mode of the file used to store logs
# Only if log_to_file is set to true
[log_file_mode: <string> | default = "0600"]

# Group to use for log resources
# Only if log_to_file is set to true
[log_group: <string> | default = "root"]

# Log level to use for process mimir
[log_level: <string> | default = "info"]

# Owner to use for log resources
# Only if log_to_file is set to true
[log_owner: <string> | default = "root"]

# Should log be kept in journald or sent to a dedicated file
[log_to_file: <boolean> | default = false]

# Command use to validate configuration
[validate_cmd: <string> | default = "/usr/local/bin/mimir --modules=true"]

# Command use to restart/reload process
[restart_cmd: <string> | default = "/bin/systemctl reload mimir"]

# Should the process be restarted on configuration changes
[restart_on_change: <boolean> | default = false]

# List of systemd parameters to override
[systemd_overrides: <hash> | default = {'Service' => {'LimitNOFILE' => '1048576'}}]

References

Deploy Grafana Mimir with Jsonnet and Tanka

Wed, 03 Jun 2026 09:01:40 +0200

Deploy Grafana Mimir with Jsonnet and Tanka

Grafana Labs publishes Jsonnet files that you can use to deploy Grafana Mimir in microservices mode. Jsonnet files are located in the Mimir repository.

Set up Mimir on Grafana Labs

Grafana Mimir hardware requirements

Grafana Mimir hardware requirements

Compute

Disk storage

Disk requirements

Unsupported filesystems

Network

Object storage

Migrate to Grafana Mimir

Migrate to Grafana Mimir

Deploy Mimir with Helm

Deploy Mimir with Helm

Use ingest storage

Example values files

See also

Deploy Grafana Mimir with Puppet

Deploy Grafana Mimir with Puppet

Start using this module

Install and configure Mimir

Configuration parameters

References

Deploy Grafana Mimir with Jsonnet and Tanka

Deploy Grafana Mimir with Jsonnet and Tanka