Skip to main content
Version: 10.3

Ports

Open the following ports on the servers. The same ports must be specified in the ports.yml file later during installation.

The following table lists the default ports for each application component.

Access types:

  • External: used by end users
  • Internal: used for internal application communication

Keep all Work.AI environment servers in the same subnet. Restrict external access to this subnet by closing all ports from outside the subnet and allow unrestricted internal access between servers by keeping all ports open within the subnet. This approach enables product components located on different hosts to communicate with each other. To ensure network security, consider using firewall rules to block requests from outside the subnet where your environment is located.

To ensure that end users can interact with the Product, open specific ports externally on the Master server. By default, the ports are 80 and 443, but you can change them if needed. In high-availability mode, 80 and 443 must be open only on the load balancer.

This is the easiest and most secure way to protect your data and application from the network perspective.

The table below describes which ports are used on which servers for communication between components.

If you decide to configure firewall rules granularly for each network interface on each server by using IPtables or FirewallD, there is a high risk of human error caused by multiple rules for each host. It is easy to forget to open a specific port from a specific server to another specific server. In this case, installation may fail at unpredictable points that are difficult to troubleshoot, or the Product may not work correctly. That is why the recommended approach to securing your network is to use a dedicated secured subnet with open internal communication and closed external requests.

Default ports

See tab Basic setup or HA setup and update from earlier versions, depending on your installation method. On the specified servers (column Destination server), open the ports (column Destination Port) to the required servers (column Source server). If the Source server cell contains "All servers", you must open the corresponding port both ways: to and from ALL existing servers, including the destination server.

In case of the basic installation, open the following ports:

Source serverDestination serverDestination portService nameAccess typeDescription
All serversMaster80, 443/tcpHAProxyexternalThe main product port for accessing from the browser
2181/tcpZooKeeperinternalListens to 0.0.0.0 (ZK). ZooKeeper port.
2281/tcpZooKeeperinternalListens to 0.0.0.0 (ZK). ZooKeeper secure port.
4567-4571/tcp, 15072-15073/tcplogstashinternalListens to 0.0.0.0 (HTTPS). Logstash port.
5672/tcpRabbitMQinternalListens to 0.0.0.0 (TLS ecrtypted AMQP). RabbitMQ message queue port.
8202/tcpHAProxyinternalListens to 0.0.0.0 (TCP). HAProxy and Load Balancer port for Vault.
MSSQL1433MS SQLinternalMS SQL Server port.
Agent22/tcpSSHinternalSSH port listening on all Linux servers. Used by the installer in a Quick installation.
5051/tcpmesos-slaveinternalListens to 0.0.0.0 (HTTPS). Mesos Agent port.
40000-44000/tcpmesos-slaveinternalListens on 0.0.0.0 (HTTPS). Port range on Mesos Agent(s) used for communication with Mesos Master.
RPA5986/tcpWinRMinternalWindows Remote Management (WinRM) HTTPS port listening to Windows servers. Used by the installer in a Quick installation.
8443/tcpNginxinternalNginx HTTPS port. Used to proxy requests from HAProxy (from lb_port_ssl) to destination product components.
Superset22/tcpSSHinternalSSH port listening on all Linux servers. Used by the installer in a Quick installation.
8443/tcpNginxinternalNginx HTTPS port. Used to proxy requests from HAProxy (from lb_port_ssl) to destination product components.
AgentMaster111/tcp+udpNFSinternalDon't change it. NFS portmap TCP/UDP port. Used by the internal NFS server only in the non-HA mode and only when nfs_enabled: True.
2049/tcp+udpNFSinternalDon't change it. NFS daemon TCP/UDP port. Used by the internal NFS server only in the non-HA mode and only when nfs_enabled: True.
5050/tcpmesos-masterinternalListens to 0.0.0.0 (HTTPS). Mesos Master port.
8480/tcpNginxinternalDon't change it. Listens to 0.0.0.0 (HTTPS). Marathon port.
9073/tcpNginxinternalListens to 0.0.0.0 (HTTPS). Nginx port. Proxies requests to ags_port.
9702/tcphazelcastinternalDon't change it. Listens to 0.0.0.0 (HTTPS). Hazelcast server port.
20048/tcp+udpNFSinternalDon't change it. NFS mountd TCP/UDP port. Used by the internal NFS server only in the non-HA mode and only when nfs_enabled: True.
Agent5051/tcpmesos-slaveinternalListens to 0.0.0.0 (HTTPS). Mesos Agent port.
40000-44000/tcpmesos-slaveinternalListens on 0.0.0.0 (HTTPS). Port range on Mesos Agent(s) used for communication with Mesos Master.
RPAMaster9073/tcpNginxinternalListens to 0.0.0.0 (HTTPS). Nginx port. Proxies requests to ags_port.
9702/tcphazelcastinternalDon't change it. Listens to 0.0.0.0 (HTTPS). Hazelcast server port.
2379/tcpetcdinternalListens to 0.0.0.0 (HTTPS). ETCD server port.
SupersetSuperset9089/tcpPodmaninternalAccess to the Superset container from Nginx.
6379/tcpPodmaninternalAccess to the Redis container from the Superset container.

Ports for Mesos and Marathon Web UI

If you plan to access the Marathon web interface (for example, to monitor workers or view their configuration), open the following ports for external access. For security reasons, you can allow access only from particular IP addresses, for example, developer, or operator workstations, using this feature:

Source serverDestination serverDestination portService nameAccess type
Operators workstationsMaster server(s)8480mesos-masterexternal

Ports for external NFS server

If you plan to use an external NFS server, open the following ports in addition to the ones mentioned earlier:

Source serverDestination serverDestination portService nameAccess type
All serversNFS111/tcp+udpNFSinternal
All serversNFS2049/tcp+udpNFSinternal
info

In case you have MOUNTD_PORT, STATD_PORT, LOCKD_TCPPORT, and LOCKD_UDPPORT statically defined on your NFS server, make sure that traffic is allowed from all Work.AI servers towards these ports on the NFS server. Otherwise, ensure that traffic is permitted from all Work.AI servers towards the port range 1024 - 65535 on the NFS server.

WorkFusion URLs

Master (in non-HA mode) or LoadBalancer (in HA mode):

  • https://workfusion_lb_hostname: Control Tower
  • https://workspace_lb_hostname/workspace: Workspace
  • https://kibana_lb_hostname/kibana/: Kibana
  • https://s3_lb_hostname: S3
  • https://nexus_lb_hostname/nexus: Nexus
  • https://rabbitmq_lb_hostname: RabbitMQ API Web UI
  • https://mesos_lb_hostname: Mesos
  • https://marathon_lb_hostname: Marathon
  • https://ocr_lb_hostname: OCR REST API
  • https://keycloak_lb_hostname: Keycloak web interface