Service list
This table describes the services that your system runs. Each service runs within its own Docker container. For each service, the table lists:
- RAM needed per instance: The amount of RAM that, by default, the service needs on each instance on which it's deployed. For all services except for System services, this value is also the default Docker Container Memory value for the service.
-
Number of instances: Shows both:
- The required number of instances on which a service must run for the system to function properly.
- The recommended number of instances that you should run a service on. These are recommended minimums; if your system includes more instances, you should take advantage of them by running services on them.
- Service unit cost per instance: The number of service units that it costs to run the service on one instance. This cost indicates how computationally expensive one service is compared to another.
- Whether the service is persistent (that is, it must run on a specific instance) or supports floating (that is, it can run on any instance).
- Whether the service has a single type or multiple.
Service name and description |
Service properties | |
The services perform functions related to the system's supported use cases. You can move, scale, and reconfigure these services. | ||
Admin-App Runs the Admin App. |
RAM needed per instance |
N/A |
Number of instances |
N/A | |
Service unit cost per instance |
10 | |
Persistent or floating |
Persistent | |
Supports volume configuration |
Yes | |
Single or multiple types |
Single | |
Cluster-Coordination Mesos (master) - https://mesos.apache.org Hardware resource management solution for distributed systems. How it's used Manages hardware resource allocation. |
RAM needed per instance |
N/A |
Number of instances |
N/A | |
Service unit cost per instance |
1 | |
Persistent or floating |
Persistent | |
Supports volume configuration |
No | |
Single or multiple types |
Single | |
Cluster-Worker Mesos (slave) - https://mesos.apache.org Hardware resource management solution for distributed systems. How it's used Receives and performs work from other services. Note: Though the Cluster-Worker service has a low service unit cost, it can at times appear to be using a large amount of CPU resources. When other services use Cluster-Worker to perform their work, Cluster-Worker reflects the CPU usage of those services. |
RAM needed per instance |
N/A |
Number of instances |
N/A | |
Service unit cost per instance |
5 | |
Persistent or floating |
Persistent | |
Supports volume configuration |
Yes | |
Single or multiple types |
Single | |
Clustered-File-System https://hortonworks.com/apache/hdfsHadoop Distributed File System. A distributed file system used for storing data. Important: The Clustered-File-System service is offered as a technology preview. Do not use it on a production system |
RAM needed per instance |
Data Node: 751 MB Name Node: 1024 MB Journal Node: 256 MB |
Number of instances | ||
Service unit cost per instance |
Data Node: 5 Name Node: 10 Journal Node: 1 | |
Persistent or floating |
Persistent | |
Supports volume configuration |
Yes | |
Single or multiple types |
Multiple:
| |
Dashboard https://www.elastic.co/products/kibana Visualizes information stored in Elasticsearch indexes. How it's used Powers the advanced Dashboard Management service. Note: This service is in the Unconfigured state by default. |
RAM needed per instance |
300 MB |
Number of instances |
Required: 0 Optimal: 2 | |
Service unit cost per instance |
5 | |
Persistent or floating |
Persistent | |
Supports volume configuration |
Yes | |
Single or multiple types |
Single | |
Database http://cassandra.apache.org/Decentralized database that can be scaled across large numbers of hardware nodes. How it's used Stores system configuration data. Also stores document discovery and failure data for workflow tasks. |
RAM needed per instance |
2.4 GB |
Number of instances |
Required: 1 Optimal: 3 | |
Service unit cost per instance |
10 | |
Persistent or floating |
Persistent | |
Supports volume configuration |
Yes | |
Single or multiple types |
Single | |
Index http://lucene.apache.org/solr/Data indexing and search platform. How it's used The search engine that manages all internal search indexes. |
RAM needed per instance |
2 GB |
Number of instances |
Required: 0 Optimal: 3 Notes:
| |
Service unit cost per instance |
25 | |
Persistent or floating |
Persistent | |
Supports volume configuration |
Yes | |
Single or multiple types |
Single | |
Logging https://www.elastic.co/products/logstashCollection engine for event data. Can perform transformations on the data it collects and then send that data to a number of outputs. How it's used Transports system logs and metrics data to the Metrics service. |
RAM needed per instance |
700 MB |
Number of instances |
Required: 1 Optimal: 1 | |
Service unit cost per instance |
10 | |
Persistent or floating |
Floating | |
Supports volume configuration |
Yes | |
Single or multiple types |
Single | |
Message Queue https://kafka.apache.org/Stream processing platform for handling real-time data streams. How it's used Facilitates communication between instances. |
RAM needed per instance |
2 GB |
Number of instances |
Required: 1 Optimal: 3 | |
Service unit cost per instance |
5 | |
Persistent or floating |
Persistent | |
Supports volume configuration |
Yes | |
Single or multiple types |
Single | |
Metrics https://www.elastic.co/Data indexing and search platform. How it's used Stores and manages:
The service maintains this information in a number of internally-managed Metrics indexes. |
RAM needed per instance |
2000 MB |
Number of instances |
Required: 1 Optimal: 3 | |
Service unit cost per instance |
25 | |
Persistent or floating |
Persistent | |
Supports volume configuration |
Yes | |
Single or multiple types |
Single | |
Monitor-App Powers the Monitor App. |
RAM needed per instance |
556 MB |
Number of instances | Required: 0 Optimal: 1 Note: Scaling the Monitor-App service does not affect any of the workflows that collect data from the systems you are monitoring. For example, if you scale the service to run on 0 instances, users cannot access the Monitor App, but HCI will continue to collect data. | |
Service unit cost per instance |
10 | |
Persistent or floating |
Floating | |
Supports volume configuration |
Yes | |
Single or multiple types |
Single | |
Network-Proxy HAProxy - https://haproxy.org Load balancer for TCP and HTTP-based applications. How it's used Maps network requests to the instances where the applicable services are located. |
RAM needed per instance |
N/A |
Number of instances |
N/A | |
Service unit cost per instance |
1 | |
Persistent or floating |
Persistent | |
Supports volume configuration |
Yes | |
Single or multiple types |
Single | |
Scheduling https://mesos.github.io/chronos/Job scheduler for Apache Mesos. How it's used Schedules workflow tasks. |
RAM needed per instance |
712 MB |
Number of instances |
Required: 1 Optimal: 1 | |
Service unit cost per instance |
1 | |
Persistent or floating |
Floating | |
Supports volume configuration |
Yes | |
Single or multiple types |
Single | |
Search-App Powers the Search App |
RAM needed per instance |
556 MB |
Number of instances | Required: 0 Optimal: 2 Note: No instances are required to run this service, but without at least one, the Search App is unavailable. | |
Service unit cost per instance |
10 | |
Persistent or floating |
Persistent | |
Supports volume configuration |
Yes | |
Single or multiple types |
Single | |
Sentinel Runs internal system processes and monitors the health of the other services. |
RAM needed per instance |
N/A |
Number of instances |
N/A | |
Service unit cost per instance |
5 | |
Persistent or floating |
Persistent | |
Supports volume configuration |
Yes | |
Single or multiple types |
Single | |
Service-Deployment Marathon - https://mesosphere.github.io/marathon/ Orchestration platform for Mesos applications. How it's used Handles deployment of high-level services (that is, the services that you can configure). |
RAM needed per instance |
N/A |
Number of instances |
N/A | |
Service unit cost per instance |
1 | |
Persistent or floating |
Persistent | |
Supports volume configuration |
Yes | |
Single or multiple types |
Single | |
Synchronization Apache Zookeeper - https://zookeeper.apache.org/ Coordinates configuration settings and other information between a number of distributed services. How it's used Coordinates actions and database operations across instances. |
RAM needed per instance |
N/A |
Number of instances |
N/A | |
Service unit cost per instance |
5 | |
Persistent or floating |
Persistent | |
Supports volume configuration |
Yes | |
Single or multiple types |
Single | |
Watchdog Monitors the other System Services and restarts them if necessary. Also responsible for initial system startup. |
RAM needed per instance |
N/A |
Number of instances |
N/A | |
Service unit cost per instance |
5 | |
Persistent or floating |
Persistent | |
Supports volume configuration |
Yes | |
Single or multiple types |
Single |
Clustered-File-System service considerations
The Clustered-File-System service gives storage space to other services in an HCI system.
Each instance of the Clustered-File-System service can be one of these types:
- Data Node: Stores data
- Name Node: Tracks the data stored in Data Nodes
- Journal Node: Tracks changes made by the Name Nodes
You can run multiple different Clustered-File-System service instance types on an HCI instance. For example, a single HCI instance can run both a Journal Node and a Data Node instances of the Clustered-File-System service.
You cannot run multiple Clustered-File-System service instances of the same type on an HCI instance. For example, a single HCI instance cannot run two Data Node instances.
The Clustered-File-System service can be deployed in either of these modes:
- High Availability (HA) Mode: In HA mode, the Clustered-File-System service can retain its stored data in the event that a Name Node instance fails or becomes corrupt.
- Non-HA Mode: In Non-HA mode, the Clustered-File-System service has a single Name Node instance. If this instance fails or becomes corrupt, all data stored by the Clustered-File-System service is lost.
The service's deployment mode is determined based on the number of service type instances you have when you initially deploy the service. After the deployment mode is set, you cannot change it without losing all data stored by the Clustered-File-System service.
To deploy the Clustered-File-System service in HA mode, you need at least three HCI instances.
Upon initial deployment of the service, you need to configure the service to run:
- Exactly two Name Node instances
- Exactly three Journal Node instances
You can run any number of Data Node instances. You need at least one to be able to store data. For HA mode, the recommended minimum is three.
To deploy the Clustered-File-System service in Non-HA mode, you need only one HCI instance.
Upon initial deployment of the service, you need to configure the service to run:
- Exactly one Name Node instance
- Zero Journal Node instances
You can run any number of Data Node instances. You need at least one to be able to store data.
When scaling or moving Name Node or Journal Node service instances, you need to perform only one operation at time.
For example, to move a Name Node from one HCI instance to another, you need to:
- Run a service operation to remove an instance of the Name Node service type.
- Wait for the operation to finish.
- Run a service operation to start a new instance of the Name Node service type.
- Wait for the operation to finish.
When running the Clustered-File-System service in HA mode, you need at least two Journal Node instances running. Scaling the service such that you have zero or one Journal Node instances causes the service to become unresponsive.
You can scale or move as many Data Node instances as necessary.
When you initially deploy the Clustered-File-System service, it is deployed in either HA or Non-HA mode, depending on the number of service instances types you initially configure the service to run.
You can change the Clustered-File-System deployment mode after the service is initially deployed, however this needs a complete redeployment of the service.
To switch from one deployment mode to another:
- Scale Clustered-File-System to run on zero instances in the system.
- Configure and run a service scale task to scale Clustered-File-System to the
applicable number of service type instances. ImportantName Node and Journal Node types must be scaled together in a single operation.