IBM UrbanCode Deploy includes several systems, including a server and one or more agents. You can configure multiple
topologies, including ones that use high availability,
containerized servers running in IBM Cloud Private, disaster recovery, and
the blueprint designer, to meet your needs.
The following system topologies, including diagrams, are shown. For explanation of the system
components, see the Description of
systems.
Core topology
The core installation of
IBM UrbanCode Deploy includes a server, agents, and a license server. Clients access the server through web browsers,
the REST API, or the command-line client. Agents can be installed on cloud environments, virtual
machines (VMs), containers, or physical systems; the agents are shown on cloud systems in the
following diagrams, but they can be installed on many different systems.
With this topology, the
server can create environments on clouds that use virtual system patterns (VSPs), such as IBM® Cloud Orchestrator and IBM PureApplication® System. To create environments on other clouds such as Amazon Web Services, SoftLayer®, VMware vCenter, and Microsoft Azure, you must install the blueprint design server and at least one engine, as
described in the blueprint design topologies.
For more information about
supported clouds that use VSPs, see Modeling environments for clouds that use virtual system patterns.
To install the core components, see Installing IBM UrbanCode Deploy.
Blueprint design topologies
To work
with blueprints on clouds via OpenStack Heat, including OpenStack-based clouds, Amazon Web Services,
SoftLayer, VMware vCenter, and Microsoft Azure, you must apply a topology that includes the blueprint design
server and engine. (In previous versions, these systems were part of
IBM UrbanCode Deploy with
Patterns.)
Installing a standalone engine
To deploy environments to non-OpenStack
clouds, such as Amazon Web Services, SoftLayer, VMware,
Google Cloud, and Microsoft Azure, you install the
blueprint design server and Heat engine through IBM UrbanCode Deploy. The following diagram shows a typical topology for this use case:
Extending Heat orchestration engines
To deploy environments to clouds
that are based on OpenStack, you install the blueprint design server and extend the Heat engine that
is associated to the cloud. In this case, you do not use the Heat engine that comes with IBM UrbanCode Deploy. The following diagram shows a typical topology for this use case:
Multi-region topology
If your environment has multiple security zones that are divided by firewalls, you can use agent
relays to connect agents to the server through the firewalls. For example, if your IBM UrbanCode Deploy server is
within a firewall, but your target environments are outside the firewall, the agents on those target
environments cannot connect directly to the server. In this case, you install an agent relay outside
the firewall to allow the agents to connect to the server through the firewall, as shown in the
following diagram.
To install and configure an agent relay, see Installing agent relays.
High-availability clustered topologies
High-availability
topologies use multiple servers. These servers can all be running at the same time to share the load
(as in a clustered topology), or they can be waiting for another server to fail (as in a cold
standby topology). The following diagram shows a clustered topology in which a load balancer
distributes connections to three servers. Users connect directly to the load balancer, which sends
them to an active server. Agents connect via HTTP and HTTPS to the load balancer; however, the
agents connect via JMS directly to the servers. The servers store their files on a shared database
and file system.
To configure a cluster of servers, see Setting up clusters of servers.
You can also cluster blueprint design servers
and engines. In this case, you install one or more blueprint design servers and engines, and set
them each to access the same database and shared file system. Similarly, a load balancer distributes
traffic to the blueprint design servers and engines. The following diagram shows a clustered
topology with three engines and three blueprint design servers, connecting to one or more
non-OpenStack clouds:
To connect to OpenStack-based clouds, the topology is different because the Heat engines
are installed via the OpenStack server, not via IBM UrbanCode Deploy:
To configure a cluster of blueprint design servers and engines, see Setting up clusters of blueprint design servers and Setting up clusters of engines.
Disaster recovery topologies
One way to prepare for
disaster recovery is to have a cold standby system, including a stopped server and a replicated copy
of the database and file system. The following diagram shows a simple topology with a cold standby
server and related resources on standby.
To configure a cold standby system for the server, see Adding cold standby servers.
You can also configure a disaster recovery system for the blueprint design server and
engine. The following diagram shows a disaster recovery topology for the blueprint design server and
Heat engine where the cold standby servers are located in a different data center. In this diagram,
the shared systems and services should be configured for high availability.
To configure disaster recovery for the blueprint designer, see Configuring disaster recovery for the blueprint design server.
Default ports
The following diagram shows the default port numbers that IBM UrbanCode Deploy uses for communication. Most of these ports can change depending on your choices at installation
time. The following diagram is only a summary of the defaults.
For more information about ports, see System requirements and performance considerations.
Description of systems
- IBM UrbanCode Deploy
server
- The IBM UrbanCode Deploy server
stores components, processes, and other elements with which you model automated application
deployments. You run automated deployments from the server.
- Web browsers
- Web browsers are the main way that users interact with the server and blueprint designer. The
IBM UrbanCode Deploy
browser-based GUI is a Rich Internet Application (RIA) that maintains much of its functionality in
the browser. Clients interact with RESTful (representational state transfer) services on the server
as needed.
- Command-line client
- The command-line client provides access to the server through the command line. It can automate
functions on the server, such as creating components and applications, and provides most of the
features that are found in the browser-based GUI. The command-line client is also built on top of
RESTful services. See Command-line client (CLI) reference.
- REST API
- The REST API provides access to the server through HTTP. Like the command-line client, it can
automate functions on the server, such as creating components and applications.
- The server and the blueprint design server have separate REST APIs. Each command in the server
REST API has an equivalent command in the command-line client; however, commands in the blueprint
design server API do not have command-line equivalents. See Extending product function.
- Agents
- The agents run processes on target systems. Agents can run on physical computers, virtual
systems, or cloud systems. See Agents.
- Agent relays
- An agent relay is a communication proxy for agents that are located behind a firewall or in
another network location. See Agent relays.
- Blueprint design server
- This server hosts the blueprint designer and controls access to blueprints, the files that
describe network topology for applications that you provision to different clouds. It also hosts the
cloud discovery service, which provides information about the available cloud resources to the
blueprint designer. (In previous versions, this system was the IBM UrbanCode Deploy with
Patterns design server.) For more information about using the blueprint designer, see Modeling environments for clouds through OpenStack Heat.
- Engine
- The Heat engine is an installation of the OpenStack Heat orchestration engine with IBM extensions. The engine manages cloud infrastructure,
provisioning resources from clouds, updating those resources, and deleting them. To connect to
non-OpenStack clouds, you install the Heat engine that is provided with IBM UrbanCode Deploy, which includes the extensions. To connect to OpenStack clouds, you extend an existing Heat
engine by adding the extensions to it. (In previous versions, this system was the IBM UrbanCode Deploy with
Patterns Heat engine.)
- Rational® Common Licensing
license server
- The license server provides licenses to the server. For more information on the license server,
see IBM
Rational Common Licensing.
- Keystone Identity Service
- This service provides authentication tokens to the OpenStack system. The blueprint design server
requires it. If you do not have a Keystone service, you can use one that is provided with the Heat
engine.
- Clouds
- The clouds host virtual resources. When you create environments with blueprints, the server or
engine provisions resources on the target clouds. For information about supported clouds, see Integrating with cloud systems.
- CodeStation
- CodeStation stores component versions and artifacts. It is part of the IBM UrbanCode Deploy server.