Configuration Parameters

This page contains the configuration parameters for both DC/OS Enterprise and DC/OS Open Source.

Cluster Setup

Networking

Storage Enterprise

Performance and Tuning

Security and Authentication

Enterprise

Open Source

Metrics

Parameter Descriptions

adminrouter_auth_cache_enabled Enterprise

Controls whether the Admin Router authorization cache is enabled.

• adminrouter_auth_cache_enabled: false Every authorization check Admin Router performs will load the user’s permissions from the IAM.
• adminrouter_auth_cache_enabled: true (Default) Admin Router will cache the user’s permissions for 5 seconds after performing an authorization check.

adminrouter_tls_1_0_enabled Enterprise

Indicates whether to enable TLS 1.0 in Admin Router. Changing this setting has no effect on internal Admin Router configuration on agent nodes.

• adminrouter_tls_1_0_enabled: 'true' Enable the TLS 1.0 protocol in Admin Router.
• adminrouter_tls_1_0_enabled: 'false' Disable the TLS 1.0 protocol in Admin Router. This is the default value.

You are advised not to enable TLS 1.0, as the protocol is considered insecure.

If you have already installed your cluster and would like to change this in place, you can go through an upgrade with the adminrouter_tls_1_0_enabled parameter set to the desired value.

adminrouter_tls_1_1_enabled Enterprise

Indicates whether to enable TLS 1.1 in Admin Router. Changing this setting has no effect on internal Admin Router configuration on agent nodes.

• adminrouter_tls_1_1_enabled: 'true' Enable the TLS 1.1 protocol in Admin Router. This is the default value.
• adminrouter_tls_1_1_enabled: 'false' Disable the TLS 1.1 protocol in Admin Router.

If you have already installed your cluster and would like to change this in-place, you can go through an upgrade with the adminrouter_tls_1_1_enabled parameter set to the desired value.

adminrouter_tls_1_2_enabled Enterprise

Indicates whether to enable TLS 1.2 in Admin Router. Changing this setting has no effect on internal Admin Router configuration on agent nodes.

• adminrouter_tls_1_2_enabled: 'true' Enable the TLS 1.2 protocol in Admin Router. This is the default value.
• adminrouter_tls_1_2_enabled: 'false' Disable the TLS 1.2 protocol in Admin Router.

It is advised to keep this protocol version enabled as it is the most secure widely supported TLS version.

If you have already installed your cluster and would like to change this in place, you can go through an upgrade with the adminrouter_tls_1_2_enabled parameter set to the desired value.

adminrouter_tls_cipher_suite Enterprise

Provide a custom list of TLS cipher suites. The value will be passed directly into Admin Router’s ssl_ciphers configuration directive. There is no validation of this string. Setting it incorrectly will cause DC/OS installation to fail. This configuration settings affects only Admin Routers running on DC/OS master nodes.

If not provided, the default value EECDH+AES128:RSA+AES128:EECDH+AES256:RSA+AES256:EECDH+3DES:RSA+3DES:!MD5; is used.

To validate the accuracy of the provided value, use the openssl ciphers utility and provide your own value: openssl ciphers <cipher-suites>. For a list of all available ciphers see the OpenSSL documentation.

NOTE: Due to Java jurisdiction limitations, it is not possible to install DC/OS with only AES256 cipher suites.

adminrouter_x_frame_options

Set the X-Frame-Options header value for the DC/OS UI. Default is set to DENY.

agent_list

A YAML nested list (-) of IPv4 addresses to your private agent host names.

auth_cookie_secure_flag Enterprise

Indicates whether to allow web browsers to send the DC/OS authentication cookie through a non-HTTPS connection. Because the DC/OS authentication cookie allows access to the DC/OS cluster, it should be sent over an encrypted connection.

• auth_cookie_secure_flag: false (default) Browsers will send the DC/OS authentication cookie through either an unencrypted HTTP connection or an encrypted HTTPS connection.

• auth_cookie_secure_flag: true The authentication cookie set by DC/OS will contain the Secure flag, which instructs the browser to not send the cookie over unencrypted HTTP connections. This could cause authentication to fail under the following circumstances.

  • If the security mode is permissive, the URL specifies HTTP, and the URL includes a target different from the root path (for example, http://<cluster-url>/<path>/)
  • There are proxies in between the browser and DC/OS that terminate TLS

aws_template_storage_access_key_id

The access key ID of the account owning the AWS S3 bucket.

aws_template_storage_bucket

The name of an S3 bucket to contain customized advanced AWS templates.

aws_template_storage_bucket_path

The path to a location within the S3 bucket to store template artifacts.

aws_template_storage_region_name

The region containing the S3 bucket.

aws_template_storage_secret_access_key

The secret access key of the account owning the AWS S3 bucket.

aws_template_upload

Whether to upload the customized advanced AWS templates to an S3 bucket.

bootstrap_url

(Required) The URL path for the DC/OS installer to store the customized DC/OS build files. If you are using the automated DC/OS installer, you should specify bootstrap_url: file:///opt/dcos_install_tmp unless you have moved the installer assets. By default the automated DC/OS installer places the build files in file:///opt/dcos_install_tmp.

bouncer_expiration_auth_token_days Enterprise

This parameter sets the auth token time-to-live (TTL) for Identity and Access Management. You must specify the value in Python float syntax wrapped in a YAML string. By default, the token expires after five days. For example, to set the token lifetime to half a day:

bouncer_expiration_auth_token_days: '0.5'

Small expiration periods may be harmful to DC/OS components. We recommend that the this value is set to no less than 0.25. If you wish to use a lower value, contact a Mesosphere support representative for guidance.

For more information, see the security documentation.

ca_certificate_path Enterprise

Use this to set a custom CA certificate. See Configuring a Certificate Authority documentation for details.

ca_certificate_key_path Enterprise

Use this to set a custom CA certificate. See Configuring a Certificate Authority documentation for details.

ca_certificate_chain_path Enterprise

Use this to set a custom CA certificate. See Configuring a Certificate Authority documentation for details.

ca_truststore_path Enterprise

Use this to set additional trusted CA certificates.

exhibitor_tls_required Enterprise

When true, failures during Exhibitor TLS bootstrap will prevent DC/OS from starting. By default, failures are logged and Exhibitor will fallback to insecure mode.

exhibitor_bootstrap_ca_url Enterprise

Optional parameter used for generating the TLS artifacts for the automated Exhibitor lock-down. This option should only be used when a bootstrap node is unavailable during DC/OS installation.

external_certificate_path Enterprise

Use this to set a custom external certificate. See Configuring a Custom External Certificate for details.

external_certificate_key_path Enterprise

Use this to set a custom external certificate. See Configuring a Custom External Certificate for details.

external_certificate_servernames Enterprise

Use this to set a custom external certificate. See Configuring a Custom External Certificate for details.

external_certificate_validation_disable Enterprise

Use this to set a custom external certificate. See Configuring a Custom External Certificate for details.

calico_enabled

Indicates whether to enable the Calico networking support. See the Calico documentation for more details. The default is true.

calico_network_cidr

Subnet allocated for calico. The subnet specified by calico_network_cidr MUST not overlap with those for VXLAN backends or virtual networks defined for DC/OS virtual networks.

calico_vxlan_enabled

Control, whether IP-in-IP or VXLAN mode is used for calico, by default VXLAN, is suggested to be used instead of VXLAN. calico_vxlan_enabled is supposed to set to ‘true’ for the environment that IP in IP is not supported, like Azure.

calico_ipinip_mtu

The MTU to set on the Calico IPIP tunnel device. This configuration works when calico_vxlan_enabled is set to be false. Please refer to the calico documentation for a suitable MTU configuration.

calico_vxlan_port

The UDP port used for calico VXLAN. This configuration works when calico_vxlan_enabled is set to be true.

calico_vxlan_vni

The virtual network ID used for calico VXLAN. This configuration works when calico_vxlan_enabled is set to be true.

calico_vxlan_mtu

The MTU to set on the Calico VXLAN tunnel device. This configuration works when calico_vxlan_enabled is set to be true. Please refer to the calico documentation for a suitable MTU configuration.

calico_veth_mtu

The MTU to set on the veth pair devices, e.g. both the container interface and host-end interface. Please refer to the calico documentation for a suitable MTU configuration.

cluster_docker_credentials

The dictionary of Docker credentials to pass.

• Creates a credentials file at cluster_docker_credentials_path during DC/OS install. A sysadmin can change credentials stored in that file as needed as long as cluster_docker_credentials_dcos_owned is set to 'false'. A systemctl restart dcos-mesos-slave or systemctl restart dcos-mesos-slave-public is required for changes to the file to take effect.
• You can also specify this value by using the --docker_config JSON format. You can write it as YAML in the config.yaml file and it will automatically be mapped to the JSON format for you.

Note:

• cluster_docker_credentials takes effect only when cluster_docker_credentials_enabled is set to 'true'.
• cluster_docker_credentials must not be left unset when cluster_docker_credentials_enabled is set to 'true'. At a minimum it must be set to a blank configuration: '{}'
• cluster_docker_credentials takes effect during an install only when either cluster_docker_credentials_dcos_owned or cluster_docker_credentials_write_to_etc are set to 'true'.
• cluster_docker_credentials takes effect during an upgrade only when cluster_docker_credentials_dcos_owned is set to ‘true’.

You can use the following options to further configure the Docker credentials:

For more information, see the examples and further documentation.

cluster_docker_credentials_dcos_owned

Whether DC/OS controls the contents of cluster_docker_credentials_path.

• cluster_docker_credentials_dcos_owned: 'true' This stores the Docker credentials in the same location as the DC/OS internal configuration (/opt/mesosphere). If you need to update or change the configuration, you will have to create a new DC/OS configuration and patch the cluster.
• cluster_docker_credentials_dcos_owned: 'false' This stores the Docker credentials outside of the DC/OS internal configuration (/opt/mesosphere). Operators can change the contents of the file and it will not be overridden during DC/OS installation or upgrade. See also: cluster_docker_credentials_path

cluster_docker_credentials_enabled

Whether to pass the Mesos --docker_config option containing cluster_docker_credentials to Mesos. The Docker configuration will be used with every task whether it uses Docker or UCR as its containerizer.

• cluster_docker_credentials_enabled: 'true' Pass the Mesos --docker_config option to Mesos. It will point to a file that contains the provided cluster_docker_credentials data.
• cluster_docker_credentials_enabled: 'false' Do not pass the Mesos --docker_config option to Mesos.

cluster_docker_credentials_path

The path and file to pass with --docker_config if cluster_docker_credentials_enabled is set to 'true'.

• If cluster_docker_credentials_dcos_owned: 'true' then this defaults to /opt/mesosphere/etc/docker_credentials.
• If cluster_docker_credentials_dcos_owned: 'false' then this defaults to /etc/mesosphere/docker_credentials.

cluster_docker_credentials_write_to_etc

Controls whether to create the file /etc/mesosphere/docker_credentials containing the data from cluster_docker_credentials. Setting this to 'false' can be useful if overwriting your credentials file will cause problems (for example, if it is part of a machine image or AMI).

• cluster_docker_credentials_write_to_etc takes effect only when cluster_docker_credentials_dcos_owned is set to false.
• cluster_docker_credentials_write_to_etc does not use cluster_docker_credentials_path, it only creates the file at /etc/mesosphere/docker_credentials.

cluster_docker_registry_url

The custom URL that Mesos uses to pull Docker images from. If set, it will configure the Mesos’ --docker_registry flag to the specified URL. This changes the default URL that Mesos uses for pulling Docker images. By default https://registry-1.docker.io is used. If changed from the default, you will need to import a local Catalog into your docker registry as you won’t access dockerhub to pull our images. See deploying a local Catalog and using a private docker registry for more information.

cluster_name

The name of your cluster.

cosmos_config

The dictionary of packaging configuration to pass to the DC/OS package manager. If set, the following options must also be specified.

• package_storage_uri Where to permanently store DC/OS packages. The value must be a file URL, for example, file:///var/lib/dcos/cosmos/packages.
• staged_package_storage_uri Where to temporarily store DC/OS packages while they are being added. The value must be a file URL, for example, file:///var/lib/dcos/cosmos/staged-packages.

custom_checks

Custom installation checks that are added to the default check configuration process. The configuration is used by the DC/OS Diagnostics component to perform installation and upgrade checks. These custom checks are run alongside the default pre- and post-flight checks during installation and upgrade.

• cluster_checks - This group of parameters specifies the health checks across the DC/OS cluster.

  • <check-name> - The custom name of your health check
  • description - Specify a description of the check
  • cmd - Specify an array of health check command strings
  • timeout - Specifies how long to wait, in seconds, before assuming the check failed. A check that times out is assumed to have a status of 3 (UNKNOWN)

• node_checks - This group of parameters specifies node health checks.

  • <check-name> - The custom name of your health check
  • description - Specify a description of the check
  • cmd - Specify an array of health check command strings
  • timeout - Specify how long to wait, in seconds, before assuming the check failed. A check that times out is assumed to have a status of 3 (UNKNOWN)

For more information on how these custom checks are used, see the examples and Node and Cluster Health Check documentation.

dcos_audit_logging Enterprise

Indicates whether security decisions (authentication, authorization) are logged for Mesos, Marathon, and Jobs.

• 'dcos_audit_logging': 'true' Mesos, Marathon, and Jobs are logged. This is the default value.
• 'dcos_audit_logging': 'false' Mesos, Marathon, and Jobs are not logged.

For more information, see the security documentation.

dcos_l4lb_enable_ipv6

Indicates whether layer-4 load-balancing is available for IPv6 containers.

• dcos_l4lb_enable_ipv6: 'false' Disables layer-4 load balancing for IPv6 containers. This is the default value.
• dcos_l4lb_enable_ipv6: 'true' Enables layer-4 load balancing for IPv6 containers.

NOTE: Layer-4 load balancing for IPv6 containers should be turned on with caution.

dcos_net_cluster_identity

This parameter ensures that the nodes from a cluster have a unique identifier that prevents unauthorized ‘cross-talk’ between clusters. Default value is false.

dcos_overlay_enable

Indicates whether to enable DC/OS virtual networks.

NOTE: Virtual networks require Docker version 1.12 or later, but if you are using Docker 1.12 or earlier then you must specify `dcos_overlay_enable: 'false'`. For more information, see the system requirements.

• dcos_overlay_enable: 'false' Do not enable the DC/OS virtual network.

• dcos_overlay_enable: 'true' Enable the DC/OS virtual network. This is the default value. After the virtual network is enabled, you can also specify the following parameters:

  • dcos_overlay_config_attempts Specifies how many failed configuration attempts are allowed before the overlay configuration modules stop trying to configure a virtual network.

    NOTE: The failures might be related to a malfunctioning Docker daemon.

  • dcos_overlay_mtu The maximum transmission unit (MTU) of the Virtual Ethernet (vEth) on the containers that are launched on the overlay.

  • dcos_overlay_network This group of parameters defines a virtual network for DC/OS. The default configuration of DC/OS provides a virtual network named dcos with this YAML configuration:

    dcos_overlay_network:
        vtep_subnet: 44.128.0.0/20
        vtep_mac_oui: 70:B3:D5:00:00:00
        overlays:
          - name: dcos
            subnet: 9.0.0.0/8
            prefix: 26
    

    • vtep_subnet A dedicated address space that is used for the VxLAN backend for the virtual network. This address space should not be accessible from outside the agents or master.

    • vtep_mac_oui The MAC address of the interface connecting to the virtual network in the public node. The last three bytes must be 00.

    • overlays

      • name The canonical name (see limitations for constraints on naming virtual networks).
      • subnet The subnet that is allocated to the virtual network.
      • prefix The size of the subnet that is allocated to each agent and thus defines the number of agents on which the overlay can run. The size of the subnet is carved from the overlay subnet.

For more information, see the example and documentation.

dcos_ucr_default_bridge_subnet

Takes an IPv4 subnet. The subnet is allocated to the bridge ucr-br0 created by the mesos-bridge CNI network. The mesos-bridge CNI network represents the network that is used to launch UCR containers when bridge-mode networking is selected for UCR containers.

The bridge-mode networking for UCR is identical to bridge mode networking for Docker and hence ucr-br0 plays the same role as docker0 bridge for Docker bridge-mode networking.

The only constraint in selecting an IPv4 subnet for dcos_ucr_default_bridge_subnet is that the subnet should not be used on the network to which the agents are connected. In other words, this subnet should be accessible from only within an agent.

dns_bind_ip_blacklist

A list of IP addresses that DC/OS DNS resolvers cannot bind to.

dns_forward_zones

A list of DNS zones, IP addresses, and ports that configure custom forwarding behavior of DNS queries. A DNS zone is mapped to a set of DNS resolvers.

A sample definition is as follows:

dns_forward_zones:
  a.contoso.com:
    - "1.1.1.1:53"
    - "2.2.2.2:53"
  b.contoso.com:
    - "3.3.3.3:53"
    - "4.4.4.4:53"

In the above example, a DNS query to myapp.a.contoso.com will be forwarded to 1.1.1.1:53 or 2.2.2.2:53. Likewise, a DNS query to myapp.b.contoso.com will be forwarded to 3.3.3.3:53 or 4.4.4.4:53.

dns_search

A space-separated list of domains that are tried when an unqualified domain is entered (for example, domain searches that do not contain ‘.’). The Linux implementation of /etc/resolv.conf restricts the maximum number of domains to 6 and the maximum number of characters the setting can have to 256. For more information, see man /etc/resolv.conf.

A search line with the specified contents is added to the /etc/resolv.conf file of every cluster host. search can do the same things as domain and is more extensible because multiple domains can be specified.

In this example, example.com has the public website www.example.com and all of the hosts in the datacenter have fully qualified domain names that end with dc1.example.com. One of the hosts in your datacenter has the hostname foo.dc1.example.com. If dns_search is set to ‘dc1.example.com example.com’, then every DC/OS host which does a name lookup of foo will get the A record for foo.dc1.example.com. If a machine looks up www, first www.dc1.example.com would be checked, but it does not exist, so the search would try the next domain, lookup www.example.com, find an A record, and then return it.

dns_search: dc1.example.com dc1.example.com example.com dc1.example.com dc2.example.com example.com

docker_remove_delay

The amount of time to wait before removing docker containers (i.e., docker rm) after Mesos regards the container as TERMINATED (for example, 3days, 2weeks, etc). This only applies for the Docker Containerizer. It is recommended that you accept the default value 1 hour.

enable_docker_gc

Indicates whether to run the docker-gc script, a simple Docker container and image garbage collection script, once every hour to clean up stray Docker containers. You can configure the runtime behavior by using the /etc/ config. For more information, see the documentation.

• enable_docker_gc: 'true' Run the docker-gc scripts once every hour. This is the default value for cloud template installations.
• enable_docker_gc: 'false' Do not run the docker-gc scripts once every hour. This is the default value for on-prem installations.

enable_mesos_input_plugin

Indicates whether to enable Telegraf’s Mesos input plugin to collect Mesos metrics from Mesos masters and agents. Options: true or false. Default value is true. For more information, see the documentation.

exhibitor_storage_backend

The type of storage backend to use for Exhibitor. You can use internal DC/OS storage (static) or specify an external storage system (ZooKeeper, aws_s3, and Azure) for configuring and orchestrating ZooKeeper with Exhibitor on the master nodes. Exhibitor automatically configures your ZooKeeper installation on the master nodes during your DC/OS installation.

• exhibitor_storage_backend: static The Exhibitor storage backend is managed internally within your cluster.

NOTE: If master_discovery is set to master_http_loadbalancer, then exhibitor_storage_backend cannot be set to static.

• exhibitor_storage_backend: zookeeper The ZooKeeper instance for shared storage. If you use a ZooKeeper instance to bootstrap Exhibitor, this ZooKeeper instance must be separate from your DC/OS cluster. You must have at least 3 ZooKeeper instances running at all times for high availability. If you specify zookeeper, you must also specify these parameters.

  • exhibitor_zk_hosts A comma-separated list (<ZK_IP>:<ZK_PORT>, <ZK_IP>:<ZK_PORT>, <ZK_IP:ZK_PORT>) of one or more ZooKeeper node IP and port addresses to use for configuring the internal Exhibitor instances. Exhibitor uses this ZooKeeper cluster to orchestrate its configuration. Multiple ZooKeeper instances are recommended for failover in production environments.
  • exhibitor_zk_path The file path that Exhibitor uses to store data.

• exhibitor_storage_backend: aws_s3 The Amazon Simple Storage Service (S3) bucket for shared storage. If you specify aws_s3, you must also specify these parameters:

  • aws_access_key_id The AWS key ID.

  • aws_region The AWS region for your S3 bucket.

  • aws_secret_access_key The AWS secret access key.

  • exhibitor_explicit_keys Indicates whether you are using AWS API keys to grant Exhibitor access to S3.

    • exhibitor_explicit_keys: 'true' If you are using AWS API keys to manually grant Exhibitor access.
    • exhibitor_explicit_keys: 'false' If you’re using an AWS Identity and Access Management (IAM) instance policy to grant Exhibitor access to S3.

  • s3_bucket The name of your S3 bucket.

  • s3_prefix The S3 prefix to be used within your S3 bucket to be used by Exhibitor.

    NOTE: AWS EC2 Classic is not supported.

• exhibitor_storage_backend: azure An Azure Storage Account for shared storage. The data will be stored under the container named dcos-exhibitor. If you specify azure, you must also specify these parameters:

  • exhibitor_azure_account_name The Azure Storage Account Name.
  • exhibitor_azure_account_key The secret key to access the Azure Storage Account.
  • exhibitor_azure_prefix The blob prefix to be used within your Storage Account to be used by Exhibitor.

enable_gpu_isolation

Indicates whether to enable GPU support in DC/OS.

• enable_gpu_isolation: 'true' Any GPUs that are installed in DC/OS will be automatically discovered and available as consumable resources for DC/OS tasks. This is the default value.
• enable_gpu_isolation: 'false' GPUs are not available for use in the cluster.

For more information, see the GPU documentation.

enable_ipv6

• enable_ipv6: 'true': Enables IPv6 networking in DC/OS. This is the default value.
• enable_ipv6: 'false': Disables IPv6 networking in DC/OS.

Currently, IPv6 networks are supported only for Docker containers. Setting this flag to true will allow the following features to be enabled:

• Users can create IPv6 DC/OS overlay networks. This will work only for Docker containers.
• Service discovery for IPv6 containers will be available.
• Layer-4 load-balancing will be available for IPv6 Docker containers if dcos_l4lb_enable_ipv6 is set to true.

fault_domain_enabled Enterprise

By default, fault domain awareness is enabled and the installer will expect input for zones and regions from a fault detect script. To override this feature, add this parameter set to false, and no fault domain information will be expected or used.

feature_dcos_storage_enabled Enterprise

Enables advanced storage features in DC/OS including CSI support for Mesos, and support for pre-installed CSI device plugins.

• feature_dcos_storage_enabled: 'false' Disables CSI support in DC/OS.
• feature_dcos_storage_enabled: 'true' Enables CSI support in DC/OS. This is the default value and it is necessary to use the DC/OS Storage Service (DSS)

gc_delay

The maximum amount of time to wait before cleaning up the executor directories. It is recommended that you accept the default value of two days.

gpus_are_scarce

Indicates whether to treat GPUs as a scarce resource in the cluster.

• gpus_are_scarce: 'true' Treat GPUs as a scarce resource. This reserves the GPUs exclusively for services that opt-in to consume GPUs via the Mesos GPU_RESOURCES framework capability. This is the default value.
• gpus_are_scarce: 'false' Treat GPUs like any other resource. GPUs will be offered indiscriminately to all frameworks, regardless of whether they use the Mesos GPU_RESOURCES framework capability or not.

iam_ldap_sync_interval Enterprise

Interval in seconds between LDAP synchronization operations.

ip_detect_public_filename

The path to a file (/genconf/ip-detect-public) on your bootstrap node that contains a shell script to map internal IPs to a public IP. For example:

#!/bin/sh
set -o nounset -o errexit

curl -fsSL https://ipinfo.io/ip

license_key_contents Enterprise

By default, the installer looks for the license key to reside in the genconf/license.txt file. The parameter license_key_contents is only used when an user wants to override this behavior and directly specify the license in the config.yaml. A license contains the maximum number of nodes attached to a cluster at any given time and the start and end date of the license, and is required for enterprise users. See these pages on licenses for more information on working with your license.

Example: license_key_contents: verylongstringofchars.

log_directory

The path to the installer host logs from the SSH processes. By default, this is set to /genconf/logs. In most cases this should not be changed because /genconf is local to the container that is running the installer, and is a mounted volume.

log_offers

Indicates whether the leading Mesos master should log the offers sent to schedulers.

• log_offers: true Enable Mesos offer logging for your cluster. This is the default value.
• log_offers: false Disable Mesos offer logging for your cluster.

marathon_gpu_scheduling_behavior

Indicates whether Marathon will schedule non-GPU tasks on nodes with an available GPU. Default is restricted, set to undefined to allow non-GPU tasks to be scheduled on agents with an available GPU. See also: metronome_gpu_scheduling_behavior

master_discovery

(Required) The Mesos master discovery method. The available options are static or master_http_loadbalancer.

• master_discovery: static Specifies that Mesos agents are used to discover the masters by giving each agent a static list of master IPs. The masters must not change IP addresses, and if a master is replaced, the new master must take the old master’s IP address.

  NOTE: On platforms like AWS where internal IPs are typically allocated dynamically, you should not use a static master list. If a master instance were to terminate for any reason, it could lead to cluster instability. Only use a static master list when you can ensure that the master IPs do not change.

  If you specify static, you must also specify this parameter:

  • master_list A YAML nested list (-) of static master IP addresses.

• master_discovery: master_http_loadbalancer The set of masters has an HTTP load balancer in front of them. The agent nodes will know the address of the load balancer. They use the load balancer to access Exhibitor on the masters to get the full list of master IPs. If you specify master_http_load_balancer, you must also specify these parameters:

  • exhibitor_address (Required) The address (preferably an IP address) of the load balancer in front of the masters. If you need to replace your masters, this address becomes the static address that agents can use to find the new master. For DC/OS Enterprise, this address is included in DC/OS certificates. The load balancer must accept traffic on ports 443, 2181, 5050, and 8181. If the cluster is running in permissive security mode, the load balancer may also accept traffic on port 80 and 8080 for non-SSL HTTP access to services in the cluster.

    NOTE: Access to the cluster over port 80 and 8080 is insecure.

    The traffic must also be forwarded to 443, 2181, 5050, and 8181 on the master. For example, Mesos port 5050 on the load balancer should forward to port 5050 on the master. The master should forward any new connections via round robin, and should avoid machines that do not respond to requests on Mesos port 5050 to ensure the master is up. For more information on security modes, check security modes documentation.

    NOTE: The internal load balancer must work in TCP mode, without any TLS termination.

  • num_masters (Required) The number of Mesos masters in your DC/OS cluster. It cannot be changed later. The number of masters behind the load balancer must never be greater than this number, though it can be lower during failures.

    NOTE: If master_discovery is set to master_http_loadbalancer, then exhibitor-storage-backend cannot be set to static. On platforms like AWS where internal IPs are allocated dynamically, you should not use a static master list. If a master instance were to terminate for any reason, it could lead to cluster instability. It is recommended to use aws_s3 for the exhibitor storage backend, since we can rely on S3 to manage quorum size when the master nodes are unavailable.

master_dns_bindall

Indicates whether the master DNS port is open. An open master DNS port listens publicly on the masters. If you are upgrading, set this parameter to true.

• master_dns_bindall: 'true' The master DNS port is open. This is the default value.
• master_dns_bindall: 'false' The master DNS port is closed.

master_external_loadbalancer Enterprise

The DNS name or IP address for the load balancer. If specified, this is included as subject alternative name in the DC/OS certificate of the Admin Router on the master nodes.

mesos_agent_work_dir Open Source

The location of the Mesos work directory on agent and public agent nodes. This defines the work_dir parameter for Mesos agents in the cluster. The default is /var/lib/mesos/slave. For details, see Mesos documentation.

mesos_container_log_sink

DC/OS can send copies of task logs to a host’s journald; however, this is disabled by default because of a known performance problem in systemd which was fixed on July 31, 2017 and included in systemd v235. However, some Linux distributions do not use this version of systemd; for example, CentOS 7.4 as of July 2018 uses systemd v219. Thus DC/OS by default does not have logrotate and journald enabled. If this feature is desired, ensure a modern version of systemd is in use in order to avoid this systemd performance problem.

The log manager for containers (tasks). The options are:

• 'journald' - send task logs only to journald
• 'logrotate' - send task logs only to the file system (i.e. a stdout/err file)
• 'journald+logrotate' - send logs to both journald and the file system

The default is logrotate. Due to performance issues, journald is not recommended. For details, see Logging Reference.

mesos_cgroups_enable_cfs

Determines whether or not CPU quota will be enforced for tasks which do not specify a CPU limit. This parameter is true by default, which means that when a DC/OS service is launched which specifies some CPU resources but does not specify a CPU limit, the service’s CPU limit will be implicitly set to the value of the CPU resources, thus constraining the service to be scheduled on the agent’s CPU no more than allowed by the implicit limit. When this parameter is false, DC/OS services which do not explicitly set CPU limits will not have their CPU usage constrained at all.

mesos_dns_set_truncate_bit

Indicates whether Mesos-DNS sets the truncate bit if the response is too large to fit in a single packet.

• mesos_dns_set_truncate_bit: 'true' Mesos-DNS sets the truncate bit if the response is too large to fit in a single packet and is truncated. This is the default behavior and complies with RFC7766.
• mesos_dns_set_truncate_bit: 'false' Mesos-DNS does not set the truncate bit if the response is too large to fit in a single packet. If you know your applications crash when resolving truncated DNS responses over TCP, or for performance reasons you want to avoid receiving the complete set of DNS records in response to your DNS requests, you should set this option to false and note that the DNS responses you receive from Mesos-DNS may be missing entries that were silently discarded. This means that truncated DNS responses will appear complete even though they are not and therefore will not trigger a retry over TCP. This behavior does not conform to RFC7766.

For more information regarding truncated DNS responses and retrying over TCP see RFC7766 - DNS Transport over TCP - Implementation Requirements.

mesos_http_executor_domain_sockets

Configures whether executors should communicate with the agent over domain sockets instead of local connections.

Mesos containerizer executors typically communicate with the agent over its v1 HTTP API over a local connection. This makes it hard to firewall the agent.

With this flag set to true, the Mesos agent will additionally listen on a domain socket, and executors started with that Mesos agent configuration active will connect over that domain socket instead.

This flag can be set to true or false; the default value is true.

mesos_master_work_dir Open Source

The location of the Mesos work directory on master nodes. This defines the work_dir parameter for Mesos masters in the cluster. The default is /var/lib/dcos/mesos/master. For details, see Mesos documentation.

mesos_max_completed_tasks_per_framework

The number of completed tasks for each framework that the Mesos master will retain in memory. In clusters with a large number of long-running frameworks, retaining too many completed tasks can cause memory issues on the master. If this parameter is not specified, the default Mesos value of 1000 is used.

mesos_seccomp_enabled

Indicates whether to enable Seccomp support for UCR containers.

• mesos_seccomp_enabled: 'true' Enables Seccomp isolator on Mesos agents. Seccomp isolator is used to set up a Seccomp profile for UCR containers.
• mesos_seccomp_enabled: 'false' Seccomp are not available for use in the cluster. This is the default value.

For more information, see the Seccomp documentation.

NOTE: DC/OS provides a default Seccomp profile, which can be enabled for UCR containers via the mesos-seccomp-profile-name option.

mesos_seccomp_profile_name

Specifies the name of the default Seccomp profile which is applied cluster-wide for UCR containers. If unset, a Seccomp profile is not applied by default. If you set this configuration option to default.json, Mesos agents will use built-in Seccomp profile. This profile is a slightly modified version of a Docker default profile. It can be found in /opt/mesosphere/etc/dcos/mesos/seccomp. It is highly recommended to use the built-in default Seccomp profile.

NOTE: This option requires the mesos_seccomp_enable option to be turned on.

metronome_gpu_scheduling_behavior

Indicates whether Metronome will schedule non-GPU tasks on nodes with an available GPU. Default is restricted, set to undefined to allow non-GPU tasks to be scheduled on agents with an available GPU. See also: marathon_gpu_scheduling_behavior

network_cni_root_dir_persist

Specifies whether to make the CNI root directory persistent during a host reboot. The default value is false. If you set this configuration option to true, the CNI root directory is created under work dir. Setting this option to true enables the CNI isolator to do proper cleanup after rebooting a host node.

NOTE: The host must be rebooted for this flag to take effect.

oauth_enabled Open Source

Indicates whether to enable authentication for your cluster.

• oauth_enabled: true Enable authentication for your cluster. This is the default value.
• oauth_enabled: false Disable authentication for your cluster.

If you have already installed your cluster and would like to disable this in place, you can go through an upgrade with the same parameter set.

permissions_cache_ttl_seconds Enterprise

The maximum number of seconds for permission changes to propagate through the entire system. Increasing this value may reduce load on the IAM by increasing the use of caches by various authorizers.

platform

The infrastructure platform. The value is optional, free-form with no content validation, and used for telemetry only. Supply an appropriate value to help inform DC/OS platform prioritization decisions. Example values: aws, azure, oneview, openstack, vsphere, vagrant-virtualbox, onprem (default).

process_timeout

The allowable amount of time, in seconds, for an action to begin after the process forks. This parameter is not the complete process time. The default value is 120 seconds. On slower networks, consider increasing to process_timeout: 600.

public_agent_list

A YAML nested list (-) of IPv4 addresses to your public agent host names.

resolvers

A YAML nested list (-) of DNS resolvers for your DC/OS cluster nodes. You can specify a maximum of 3 resolvers. Set this parameter to the most authoritative nameservers that you have.

• If you want to resolve internal hostnames, set it to a nameserver that can resolve them.

• If you do not have internal hostnames to resolve, you can set this to a public nameserver like Google or AWS. For example, you can specify the Google Public DNS IP addresses (IPv4):

  resolvers:
  - 8.8.4.4
  - 8.8.8.8
  

• If you do not have a DNS infrastructure and do not have access to internet DNS servers, you can specify resolvers: []. By specifying this setting, all requests to non-.mesos will return an error. For more information, see the Mesos-DNS documentation.

WARNING: If you set the resolvers parameter incorrectly, you will permanently damage your configuration and have to reinstall DC/OS.

rexray_config

The REX-Ray configuration for enabling external persistent volumes in Marathon. REX-Ray is a storage orchestration engine. The following is an example configuration.

rexray_config:
    rexray:
      loglevel: info
      service: ebs
    libstorage:
      integration:
        volume:
          operations:
            unmount:
              ignoreusedcount: true
      server:
        tasks:
          logTimeout: 5m

See the DVDI volumes documentation for information on how to create your configuration.

If the rexray_config parameter is provided, its contents are used verbatim for REX-Ray’s configuration. This lets you define completely custom REX-Ray configurations which integrate with various external storage providers. However, if you upgrade your cluster to a version that includes an updated version of REX-Ray, you must ensure that your rexray_config parameter is compatible with the newer version of REX-Ray.

rexray_config_preset

If you are running your cluster on AWS, and want DC/OS to integrate with the Elastic Block Storage (EBS) without caring about the specific REX-Ray configuration, set the rexray_config_preset parameter to aws. This sets the rexray_config parameter to the default REX-Ray configuration bundled with DC/OS. This option also has the benefit of automatically upgrading your cluster’s REX-Ray configuration when you upgrade to a newer version of DC/OS that contains an updated REX-Ray version.

security Enterprise

Specify a security mode other than security: permissive (the default). The possible values follow.

• security: permissive
• security: strict

Refer to the security modes section for a detailed discussion of each parameter.

ssh_key_path

The path that the installer uses to log into the target nodes. By default this is set to /genconf/ssh_key. This parameter should not be changed because /genconf is local to the container that is running the installer, and is a mounted volume.

ssh_port

The port to SSH to, for example 22.

ssh_user

The SSH username, for example centos.

superuser_password_hash Enterprise

(Required) The hashed superuser password. The superuser_password_hash is generated by using the installer --hash-password flag. This first super user account is used to provide a method of logging into DC/OS, at which point additional administrative accounts can be added. For more information, see the security documentation.

superuser_service_account_uid

This service account authenticates using its public key specified by superuser_service_account_public_key configuration parameter.

superuser_service_account_public_key

The superuser service account public key. This additional superuser account provides a method to automate tasks against a DC/OS cluster with superuser privileges. The operator can create a RSA key-pair and supply the generated public key in X.509 PEM public key format.

superuser_username (Required) Enterprise

The user name of the superuser. This account uses the superuser_password_hash. For more information, see the security documentation.

telemetry_enabled

Indicates whether to enable sharing of anonymous data for your cluster.

• telemetry_enabled: 'true' Enable anonymous data sharing. This is the default value.
• telemetry_enabled: 'false' Disable anonymous data sharing.

If you have already installed your cluster and would like to disable this in place, you can go through an upgrade with the same parameter set.

use_proxy

Indicates whether to enable the DC/OS proxy.

• use_proxy: 'false' Do not configure DC/OS components to use a custom proxy. This is the default value.

• use_proxy: 'true' Configure DC/OS components to use a custom proxy. If you specify use_proxy: 'true', you can also specify these parameters:

  NOTE: The specified proxies must be resolvable from the provided list of resolvers.

  • http_proxy: http://<user>:<pass>@<proxy_host>:<http_proxy_port> The HTTP proxy.
  • https_proxy: https://<user>:<pass>@<proxy_host>:<https_proxy_port> The HTTPS proxy.
  • no_proxy: A YAML nested list (-) of subdomains to exclude from forwarding to the https_proxy. If the address matches one of these strings, or the host is within the domain of one of these strings, http(s) requests to that node are not proxied. For example, the no_proxy list can be a list of internal IP addresses.

  NOTE: Wildcard characters (*) are not supported.

  For more information, see the examples.

  NOTE: You should also configure an HTTP proxy for Docker

zk_super_credentials Enterprise

On DC/OS strict and permissive mode clusters the information stored in ZooKeeper is protected using access control lists (ACLs), so that a malicious user cannot connect to the ZooKeeper Quorum and directly modify service metadata. ACLs specify sets of resource IDs (RIDs) and actions that are associated with those IDs. ZooKeeper supports pluggable authentication schemes and has a few built in schemes: world, auth, digest, host, and ip.

DC/OS ZooKeeper credentials zk_super_credentials, zk_master_credentials, and zk_agent_credentials use digest authentication, which requires a <uid>:<password> string which is then used as an ID while checking if a client can access a particular resource.

zk_super_credentials enables access to ZooKeeper’s equivalent of the root or superuser account, which has access to all resources regardless of existing ACLs. This credential allows an operator to access all the metadata stored in the ZooKeeper Quorum and is used by the DC/OS bootstrap script while initializing the cluster. Default: 'super:secret'.

To harden clusters, Mesosphere recommends that you change the defaults of all credentials to long, complex values. Once set, you can verify the settings using /opt/mesosphere/active/exhibitor/usr/zookeeper/bin/zkCli.sh available on DC/OS master nodes. By default, zkCli does not authenticate, so the nodes in the /dcos tree will not be accessible. After invoking addauth digest <zk_super_credentials> in zkCli, all the nodes in ZooKeeper will be accessible, with zk_master_credentials and zk_agent_credentials providing access to a subset of them. For example:

[zk: localhost:2181(CONNECTED) 0] addauth digest super:secret
[zk: localhost:2181(CONNECTED) 1] ls /dcos
[backup, agent, RootCA, secrets, vault, CAChainInclRoot, CAChain, CACertKeyType, ca, master]
[zk: localhost:2181(CONNECTED) 2] ls /dcos/secrets
[core, init, system, bootstrap_user, keys]

zk_master_credentials Enterprise

Credentials used by the bootstrapping processes to access the credentials of the services that will be running on the DC/OS master nodes.

zk_agent_credentials Enterprise

Credentials used by the bootstrapping processes to access the credentials of the services that will be running on the DC/OS agent nodes.