The Harbor service provisions ports to user applications. The standard router in Stackato is designed for HTTP(S) traffic only, if user applications require other protocols, consider setting up Harbor.
Information on using the Harbor port service with deployed applications is in the Port Service (Harbor) documentation.
A Harbor node is designed to sit on the edge of the cluster network (i.e. with a publicly routable network interface) in a similar fashion to the router component. This allows it to effectively proxy external and internal traffic from your app to the outside world and vice versa.
Unlike the router component which normally only requires 80/HTTP and 443/HTTPS ports to be externally visible, a Harbor node will automatically provision ports from a pre-configured range.
The harbor node learns about new routes for each provisioned service from the DEAs. When an application instance bound to a Harbor service on a DEA is terminated (stopped, failed, or scaled down), Harbor will also drop that instance from its list of backends available for the externally facing port.
A Stackato VM can be configured as a Harbor node in the same way as other data service roles. For example:
$ kato node attach -e harbor CORE_IP
The node must be routable both externally by connecting clients, and the internal DEA nodes must be contactable by the Harbor node. You must also make allowances for any firewall rules so that the provisioned port range used by Harbor is open to the external network.
The default range for these ports is '35000 - 40000', however you can view the current port range on the Harbor node by entering:
$ kato config get harbor_node port_range
For each new service provisioned, Harbor will chose a random unassigned port from this range. This range can be set in the Management Console's Settings section, or by using the kato config set ... command.
A Harbor node can run standalone, or on the same node as your router. This may be the preferred option if wish to use the same DNS name for Harbor and Router services. Otherwise, create new DNS entry for the harbor node, so that consumers of the port do not have to address it by its external IP.
If you are setting a different IP / DNS name for this port you should update the node config so that it provides these external connection details to the users service credentials.
To set the externally routable hostname:
$ kato config set harbor_node hostname_external ext-services.example.com
To set the externally routable IP:
$ kato config set harbor_node host_external 184.108.40.206
If you have problems with the Harbor service, first check the status via 'kato status'.
The Harbor service can be safely restarted; any ports and routes provisioned are automatically added and the backends are re-validated on startup:
$ kato restart harbor
Check the log files for errors and warnings. Four Harbor log files can be found in the /s/logs/ directory:
* harbor_gateway.log * harbor_node.log * harbor_proxy_connector.log * harbor_redis.log
No Stackato processes should conflict with the default Harbor port range. However, in systems with additional custom processes running, other processes may assume control of a port within Harbor's port range. This condition which will show up in harbor_proxy_connector.log.
You can avoid these conflicts by setting the Harbor port range to avoid conflicting processes. To resolve an accidental conflict, change the port used by the conflicting application and restart. Any user application using the port in conflict should be restarted to force Harbor to re-listen on that port.