.. _deploy-minio-distributed: .. _minio-mnmd: ==================================== Deploy MinIO: Multi-Node Multi-Drive ==================================== .. default-domain:: minio .. contents:: Table of Contents :local: :depth: 1 Overview -------- A distributed MinIO deployment consists of 4 or more drives/volumes managed by one or more :mc:`minio server` process, where the processes manage pooling the compute and storage resources into a single aggregated object storage resource. Each MinIO server has a complete picture of the distributed topology, such that an application can connect to any node in the deployment and perform S3 operations. Distributed deployments implicitly enable :ref:`erasure coding `, MinIO's data redundancy and availability feature that allows deployments to automatically reconstruct objects on-the-fly despite the loss of multiple drives or nodes in the cluster. Erasure coding provides object-level healing with less overhead than adjacent technologies such as RAID or replication. Depending on the configured :ref:`erasure code parity `, a distributed deployment with ``m`` servers and ``n`` disks per server can continue serving read and write operations with only ``m/2`` servers or ``m*n/2`` drives online and accessible. Distributed deployments also support the following features: - :ref:`Server-Side Object Replication ` - :ref:`Write-Once Read-Many Locking ` - :ref:`Object Versioning ` .. _deploy-minio-distributed-prereqs: Prerequisites ------------- Networking and Firewalls ~~~~~~~~~~~~~~~~~~~~~~~~ Each node should have full bidirectional network access to every other node in the deployment. For containerized or orchestrated infrastructures, this may require specific configuration of networking and routing components such as ingress or load balancers. Certain operating systems may also require setting firewall rules. For example, the following command explicitly opens the default MinIO server API port ``9000`` for servers running firewalld : .. code-block:: shell :class: copyable firewall-cmd --permanent --zone=public --add-port=9000/tcp firewall-cmd --reload All MinIO servers in the deployment *must* use the same listen port. If you set a static :ref:`MinIO Console ` port (e.g. ``:9001``) you must *also* grant access to that port to ensure connectivity from external clients. MinIO **strongly recomends** using a load balancer to manage connectivity to the cluster. The Load Balancer should use a "Least Connections" algorithm for routing requests to the MinIO deployment, since any MinIO node in the deployment can receive, route, or process client requests. The following load balancers are known to work well with MinIO: - `NGINX `__ - `HAProxy `__ Configuring firewalls or load balancers to support MinIO is out of scope for this procedure. Sequential Hostnames ~~~~~~~~~~~~~~~~~~~~ MinIO *requires* using expansion notation ``{x...y}`` to denote a sequential series of MinIO hosts when creating a server pool. MinIO therefore *requires* using sequentially-numbered hostnames to represent each :mc:`minio server` process in the deployment. Create the necessary DNS hostname mappings *prior* to starting this procedure. For example, the following hostnames would support a 4-node distributed deployment: - ``minio1.example.com`` - ``minio2.example.com`` - ``minio3.example.com`` - ``minio4.example.com`` You can specify the entire range of hostnames using the expansion notation ``minio{1...4}.example.com``. Configuring DNS to support MinIO is out of scope for this procedure. .. _deploy-minio-distributed-prereqs-storage: Local JBOD Storage with Sequential Mounts ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ .. |deployment| replace:: deployment .. include:: /includes/common-installation.rst :start-after: start-local-jbod-desc :end-before: end-local-jbod-desc .. admonition:: Network File System Volumes Break Consistency Guarantees :class: note MinIO's strict **read-after-write** and **list-after-write** consistency model requires local disk filesystems. MinIO cannot provide consistency guarantees if the underlying storage volumes are NFS or a similar network-attached storage volume. For deployments that *require* using network-attached storage, use NFSv4 for best results. Considerations -------------- Homogeneous Node Configurations ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ MinIO strongly recommends selecting substantially similar hardware configurations for all nodes in the deployment. Ensure the hardware (CPU, memory, motherboard, storage adapters) and software (operating system, kernel settings, system services) is consistent across all nodes. Deployment may exhibit unpredictable performance if nodes have heterogeneous hardware or software configurations. Workloads that benefit from storing aged data on lower-cost hardware should instead deploy a dedicated "warm" or "cold" MinIO deployment and :ref:`transition ` data to that tier. Erasure Coding Parity ~~~~~~~~~~~~~~~~~~~~~ MinIO :ref:`erasure coding ` is a data redundancy and availability feature that allows MinIO deployments to automatically reconstruct objects on-the-fly despite the loss of multiple drives or nodes in the cluster. Erasure Coding provides object-level healing with less overhead than adjacent technologies such as RAID or replication. Distributed deployments implicitly enable and rely on erasure coding for core functionality. Erasure Coding splits objects into data and parity blocks, where parity blocks support reconstruction of missing or corrupted data blocks. The number of parity blocks in a deployment controls the deployment's relative data redundancy. Higher levels of parity allow for higher tolerance of drive loss at the cost of total available storage. MinIO defaults to ``EC:4`` , or 4 parity blocks per :ref:`erasure set `. You can set a custom parity level by setting the appropriate :ref:`MinIO Storage Class environment variable `. Consider using the MinIO `Erasure Code Calculator `__ for guidance in selecting the appropriate erasure code parity level for your cluster. Capacity-Based Planning ~~~~~~~~~~~~~~~~~~~~~~~ MinIO generally recommends planning capacity such that :ref:`server pool expansion ` is only required after 2+ years of deployment uptime. For example, consider an application suite that is estimated to produce 10TB of data per year. The MinIO deployment should provide *at minimum*: ``10TB + 10TB + 10TB = 30TB`` MinIO recommends adding buffer storage to account for potential growth in stored data (e.g. 40TB of total usable storage). As a rule-of-thumb, more capacity initially is preferred over frequent just-in-time expansion to meet capacity requirements. Since MinIO :ref:`erasure coding ` requires some storage for parity, the total **raw** storage must exceed the planned **usable** capacity. Consider using the MinIO `Erasure Code Calculator `__ for guidance in planning capacity around specific erasure code settings. Recommended Operating Systems ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ .. cond:: linux This tutorial assumes all hosts running MinIO use a :ref:`recommended Linux operating system ` such as RHEL8+ or Ubuntu 18.04+. .. cond:: macos This tutorial assumes all hosts running MinIO use a non-EOL macOS version (10.14+). .. cond:: Windows This tutorial assumes all hosts running MinIO use a non-EOL Windows distribution. Support for running distributed MinIO deployments on Windows is *experimental*. Pre-Existing Data ~~~~~~~~~~~~~~~~~ When starting a new MinIO server in a distributed environment, the storage devices must not have existing data. Once you start the MinIO server, all interactions with the data must be done through the S3 API. Use the :ref:`MinIO Client `, the :ref:`MinIO Console `, or one of the MinIO :ref:`Software Development Kits ` to work with the buckets and objects. .. warning:: Modifying files on the backend drives can result in data corruption or data loss. .. _deploy-minio-distributed-baremetal: Deploy Distributed MinIO ------------------------ The following procedure creates a new distributed MinIO deployment consisting of a single :ref:`Server Pool `. All commands provided below use example values. Replace these values with those appropriate for your deployment. Review the :ref:`deploy-minio-distributed-prereqs` before starting this procedure. 1) Install the MinIO Binary on Each Node ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ .. cond:: linux .. include:: /includes/linux/common-installation.rst :start-after: start-install-minio-binary-desc :end-before: end-install-minio-binary-desc .. cond:: macos .. include:: /includes/macos/common-installation.rst :start-after: start-install-minio-binary-desc :end-before: end-install-minio-binary-desc 2) Create the ``systemd`` Service File ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ .. include:: /includes/linux/common-installation.rst :start-after: start-install-minio-systemd-desc :end-before: end-install-minio-systemd-desc 3) Create the Service Environment File ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ Create an environment file at ``/etc/default/minio``. The MinIO service uses this file as the source of all :ref:`environment variables ` used by MinIO *and* the ``minio.service`` file. The following examples assumes that: - The deployment has a single server pool consisting of four MinIO server hosts with sequential hostnames. .. code-block:: shell minio1.example.com minio3.example.com minio2.example.com minio4.example.com - All hosts have four locally-attached disks with sequential mount-points: .. code-block:: shell /mnt/disk1/minio /mnt/disk3/minio /mnt/disk2/minio /mnt/disk4/minio - The deployment has a load balancer running at ``https://minio.example.net`` that manages connections across all four MinIO hosts. Modify the example to reflect your deployment topology: .. code-block:: shell :class: copyable # Set the hosts and volumes MinIO uses at startup # The command uses MinIO expansion notation {x...y} to denote a # sequential series. # # The following example covers four MinIO hosts # with 4 drives each at the specified hostname and drive locations. # The command includes the port that each MinIO server listens on # (default 9000) MINIO_VOLUMES="https://minio{1...4}.example.net:9000/mnt/disk{1...4}/minio" # Set all MinIO server options # # The following explicitly sets the MinIO Console listen address to # port 9001 on all network interfaces. The default behavior is dynamic # port selection. MINIO_OPTS="--console-address :9001" # Set the root username. This user has unrestricted permissions to # perform S3 and administrative API operations on any resource in the # deployment. # # Defer to your organizations requirements for superadmin user name. MINIO_ROOT_USER=minioadmin # Set the root password # # Use a long, random, unique string that meets your organizations # requirements for passwords. MINIO_ROOT_PASSWORD=minio-secret-key-CHANGE-ME # Set to the URL of the load balancer for the MinIO deployment # This value *must* match across all MinIO servers. If you do # not have a load balancer, set this value to to any *one* of the # MinIO hosts in the deployment as a temporary measure. MINIO_SERVER_URL="https://minio.example.net:9000" You may specify other :ref:`environment variables ` or server commandline options as required by your deployment. All MinIO nodes in the deployment should include the same environment variables with the same values for each variable. 4) Add TLS/SSL Certificates ~~~~~~~~~~~~~~~~~~~~~~~~~~~ .. include:: /includes/common-installation.rst :start-after: start-install-minio-tls-desc :end-before: end-install-minio-tls-desc 5) Run the MinIO Server Process ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ Issue the following commands on each node in the deployment to start the MinIO service: .. include:: /includes/linux/common-installation.rst :start-after: start-install-minio-start-service-desc :end-before: end-install-minio-start-service-desc 6) Open the MinIO Console ~~~~~~~~~~~~~~~~~~~~~~~~~ .. include:: /includes/common-installation.rst :start-after: start-install-minio-console-desc :end-before: end-install-minio-console-desc 7) Next Steps ~~~~~~~~~~~~~ - Create an :ref:`alias ` for accessing the deployment using :mc:`mc`. - :ref:`Create users and policies to control access to the deployment `. .. _deploy-minio-distributed-recommendations: Deployment Recommendations -------------------------- Minimum Nodes per Deployment ~~~~~~~~~~~~~~~~~~~~~~~~~~~~ For all production deployments, MinIO recommends a *minimum* of 4 nodes per :ref:`server pool ` with 4 drives per server. With the default :ref:`erasure code parity ` setting of ``EC:4``, this topology can continue serving read and write operations despite the loss of up to 4 drives *or* one node. The minimum recommendation reflects MinIO's experience with assisting enterprise customers in deploying on a variety of IT infrastructures while maintaining the desired SLA/SLO. While MinIO may run on less than the minimum recommended topology, any potential cost savings come at the risk of decreased reliability. Server Hardware ~~~~~~~~~~~~~~~ MinIO is hardware agnostic and runs on a variety of hardware architectures ranging from ARM-based embedded systems to high-end x64 and POWER9 servers. The following recommendations match MinIO's `Reference Hardware `__ for large-scale data storage: .. list-table:: :stub-columns: 1 :widths: 20 80 :width: 100% * - Processor - Dual Intel Xeon Scalable Gold CPUs with 8 cores per socket. * - Memory - 128GB of Memory per pod * - Network - Minimum of 25GbE NIC and supporting network infrastructure between nodes. MinIO can make maximum use of drive throughput, which can fully saturate network links between MinIO nodes or clients. Large clusters may require 100GbE network infrastructure to fully utilize MinIO's per-node performance potential. * - Drives - SATA/SAS NVMe/SSD with a minimum of 8 drives per server. Drives should be :abbr:`JBOD (Just a Bunch of Disks)` arrays with no RAID or similar technologies. MinIO recommends XFS formatting for best performance. Use the same type of disk (NVMe, SSD, or HDD) with the same capacity across all nodes in the deployment. MinIO does not distinguish drive types when using the underlying storage and does not benefit from mixed storage types. Additionally. MinIO limits the size used per disk to the smallest drive in the deployment. For example, if the deployment has 15 10TB disks and 1 1TB disk, MinIO limits the per-disk capacity to 1TB. Networking ~~~~~~~~~~ MinIO recommends high speed networking to support the maximum possible throughput of the attached storage (aggregated drives, storage controllers, and PCIe busses). The following table provides general guidelines for the maximum storage throughput supported by a given NIC: .. list-table:: :header-rows: 1 :width: 100% :widths: 40 60 * - NIC bandwidth (Gbps) - Estimated Aggregated Storage Throughput (GBps) * - 10GbE - 1GBps * - 25GbE - 2.5GBps * - 50GbE - 5GBps * - 100GbE - 10GBps CPU Allocation ~~~~~~~~~~~~~~ MinIO can perform well with consumer-grade processors. MinIO can take advantage of CPUs which support AVX-512 SIMD instructions for increased performance of certain operations. MinIO benefits from allocating CPU based on the expected per-host network throughput. The following table provides general guidelines for allocating CPU for use by based on the total network bandwidth supported by the host: .. list-table:: :header-rows: 1 :width: 100% :widths: 40 60 * - Host NIC Bandwidth - Recommended Pod vCPU * - 10GbE or less - 8 vCPU per pod. * - 25GbE - 16 vCPU per pod. * - 50GbE - 32 vCPU per pod. * - 100GbE - 64 vCPU per pod. .. _minio-k8s-production-considerations-memory: Memory Allocation ~~~~~~~~~~~~~~~~~ MinIO benefits from allocating memory based on the total storage of each host. The following table provides general guidelines for allocating memory for use by MinIO server processes based on the total amount of local storage on the host: .. list-table:: :header-rows: 1 :width: 100% :widths: 40 60 * - Total Host Storage - Recommended Host Memory * - Up to 1 Tebibyte (Ti) - 8GiB * - Up to 10 Tebibyte (Ti) - 16GiB * - Up to 100 Tebibyte (Ti) - 32GiB * - Up to 1 Pebibyte (Pi) - 64GiB * - More than 1 Pebibyte (Pi) - 128GiB .. _minio-requests-per-node: Requests Per Node ~~~~~~~~~~~~~~~~~ You can calculate the maximum number of concurrent requests per host with this formula: :math:`totalRam / ramPerRequest` To calculate the amount of RAM used for each request, use this formula: :math:`((2MiB + 128KiB) * driveCount) + (2 * 10MiB) + (2 * 1 MiB)` 10MiB is the default erasure block size v1. 1 MiB is the default erasure block size v2. The following table lists the maximum concurrent requests on a node based on the number of host drives and the *free* system RAM: .. list-table:: :header-rows: 1 :width: 100% * - Number of Drives - 32 GiB of RAM - 64 GiB of RAM - 128 GiB of RAM - 256 GiB of RAM - 512 GiB of RAM * - 4 Drives - 1,074 - 2,149 - 4,297 - 8,595 - 17,190 * - 8 Drives - 840 - 1,680 - 3,361 - 6,722 - 13,443 * - 16 Drives - 585 - 1,170 - 2.341 - 4,681 - 9,362