Download latest OctopusPuppet.Cmd

Download latest master OctopusPuppet.Gui

Download latest master OctopusPuppet.Teamcity Plugin

See https://github.com/Aqovia/OctopusPuppet for information about Octopus Puppet.

Generate deployment plan

Visualize and configure deployment plan

Deploy

Run it from command line or during CI

Provides a way to visualize, orchestrate differential deployments in Octopus.

Will deploy all components that have SemVer pre-release that matches given branch. The latest component will be selected. If the target environment already has the exact component it will be skipped.

All components that do not have a matching SemVer pre-release will then match no SemVer pre-release (i.e. master). The latest component will be selected. If the target environment already has the component it will be skipped.

This is often used for testing a feature branch. Features can span multiple repositories which means multiple components.

Will deploy all components from source environment to target environment. The exact component is matched. If the target environment already has the exact component it will be skipped.

This is often used for promoting a product from one environment to another or when developers want to debug a product on their own machines.

Will redeploy all components. The exact component is matched.

This is often used for redeploying when settings/feature toggles have changed.

When running micro-services architecture one of the downsides is that there is plenty of stuff to deploy. To reduce this tax you need to ensure as much as possible is automated. Octopus goes a long way to making a great experience, but there are a few short comings.

Deployment Taxes:

• It is often difficult to see how components are dependent on each other. The ideal situation is a graphical graph which shows component deployment dependencies.
• Having one giant octopus deployment project is not feasible either. Often we don't want to deploy everything as this takes ages. The ideal situation is a differential deployment. It will only deploy what has not already been deployed.
• There appears to be no way of calling other deployments from a deployment in Octopus. This is due to Octopus will block the deployment. So you can't call the Octopus API and wait until the deployment is finished as Octopus will only start running the task once the wait is finished. The ideal situation is a way to call deployments from deployments or have something that can orchestrate this.

There is no way to show component deployment dependencies in Octopus. To fake this, add a variable called "ComponentDependencies" to a component. This variable is a json string array of all components that this component is dependent on.

Example component dependencies

['Payment Service', 'Fraud Service']

This is susceptible to people refactoring component names or deleting dependent components, but OctopusPuppet will detect when it is dependent on a component that does not exist. It will skip these dependencies and will show in red in the deployment planner.

There may be a number of unrelated products configured for an environment in Octopus. So when you want to do a branch deployment you may not care about all the other products. In order to support this, there is a concept of a component filter. This is simple a list of regular expressions to either include or exclude components based on their name.

If you exclude a component that another component is dependent on it will filter the ComponentDependencies variable as well.

This allows you do something like this:

• Create a filter for environment deployments
• Step 1 - Deploy database schema
• Step 2 - Deploy website
• Create a filter for environment refreshes
• Step 1 - Restore database
• Step 2 - Deploy database schema
• Step 3 - Deploy website

A -> B -> C
      \-> D -> E

• Step 1
• Deploy A
• Step 2
• Deploy B
• Step 3
• Deploy C
• Deploy D
• Step 4
• Deploy E

The deployment scheduler will calculate the optimum deployment schedule, by parallezing deployments when it can. It can also deal with cyclical component dependencies.

Theory: https://en.wikipedia.org/wiki/Directed_graph on how to model component dependencies https://en.wikipedia.org/wiki/Directed_acyclic_graph on how our deployment steps should be

A -> B -> C -> A
      \-> D -> E

• Step 1
• Deploy A
• Deploy B
• Deploy C
• Step 2
• Deploy D
• Step 3
• Deploy E

When the deployment scheduler detect cyclical component dependencies it will plan to deploy each of these components in parallel as there is no clear order to deploy them in. So it will deploy these components in parallel.

A -> B -> C
      \-> D -> E
Z->Y

• Product 1
• A
• B
• C
• D
• E
• Product 2
• Z
• Y

The deployment scheduler will detect product by grouping components that are strongly connected. A product is something that can be deployed without effecting another product, so its useful to calculate what a product it.

Theory: https://en.wikipedia.org/wiki/Strongly_connected_component on how to work out product.

The are two approaches to database deployments. Change sets vs declarative schema. If you use declarative schema (e.g. Redgate) then you can switch between branches without too much trouble. If you use change sets (e.g. Liquibase) to deploy database changes then you are going to have to restore the database, then apply change sets when switching branches.

Provides a way run Octopus Puppet from Teamcity so that you can orchestrate differential deployments in Octopus.

Will deploy all components that have SemVer pre-release that matches given branch. The latest component will be selected. If the target environment already has the exact component it will be skipped.

All components that do not have a matching SemVer pre-release will then match no SemVer pre-release (i.e. master). The latest component will be selected. If the target environment already has the component it will be skipped.

This is often used for testing a feature branch. Features can span multiple repositories which means multiple components.

Will deploy all components from source environment to target environment. The exact component is matched. If the target environment already has the exact component it will be skipped.

This is often used for promoting a product from one environment to another or when developers want to debug a product on their own machines.

Will redeploy all components. The exact component is matched.

This is often used for redeploying when settings/feature toggles have changed.

Often you want to orchastrate the deployment of environments during CI. The most common case is doing a branch deployment to run integration tests against it. You might also want to deploy master branch environment for overnight tests (soak/performance/huge regression test suites).

Teamcity plugin the will expose the command line version of Octopus Puppet via the Teamcity GUI