ChangeLog.md

Change Log

version v0.9.1

• Update internal references to use the new local.jmpd.in domain bypassing problems where chrome auto redirects .dev to https://.
• Update Nomad to 1.7.5

version v0.7.0

• Add capability to add custom container registries to the image cache

  Nomad and Kuberentes clusters are started in a Docker container that does not save any state to the local disk. This state includes and Docker Image cache, thefore every time an image is pulled to a new cluster it is downloaded from the internet. This can be slow and bandwidth intensive. To solve this problem Jumppad implemented a pull through cache that is used by all clusters. By default this cache supported the following registires:

  • k8s.gcr.io
  • gcr.io
  • asia.gcr.io
  • eu.gcr.io
  • us.gcr.io
  • quay.io
  • ghcr.io
  • docker.pkg.github.com

  To support custom registries Jumppad has added a new resource type container_registry. This resource type can be used to define either a local or remote registry. When a registry is defined it is added to the pull through cache and any authnetication details are added to the cache meaning you do not need to authenticate each pull on the Nomad or Kubernetes cluster. Any defined registry must be configured to use HTTPS, the image cache can not be used to pull from insecure registries.

# Define a custom registry that does not use authentication
resource "container_registry" "noauth" {
  hostname = "noauth-registry.demo.gs" // cache can not resolve local.jmpd.in dns for some reason, 
                                       // using external dns mapped to the local ip address
}

# Define a custom registry that uses authentication
resource "container_registry" "auth" {
  hostname = "auth-registry.demo.gs"
  auth {
    username = "admin"
    password = "password"
  }
}

• Add capability to add insecure registries and image cache bypass to Kubernetes and Nomad clusters.

  All images pulled to Nomad and Kubernetes clusters are pulled through the image cache. This cache is a Docker container that is automatically started by Jumppad. To disable the cache and pull images directly from the internet you can add the no_proxy parameter to the new docker config stanza. This will cause the cache to be bypassed and the image to be pulled direct from the internet.

  To support insecure registries you can add the insecure_registries parameter to the docker config stanza. This must be used in conjunction with the no_proxy parameter as the image cache does not support insecure registries.

resource "nomad_cluster" "dev" {
  client_nodes = 1

  datacenter = "dc1"

  network {
    id = variable.network_id
  }

  // add configuration to allow cache bypass and insecure registry
  config {
    docker {
      no_proxy            = ["insecure.container.local.jmpd.in"]
      insecure_registries = ["insecure.container.local.jmpd.in:5003"]
    }
  }
}

version v0.5.47

• Fix isuse where filepath.Walk does not respect symlinks
• Add ignore parameter to build resource to allow ignoring of files and folders for Docker builds.

version v0.5.43

• Add ability to set datacenter for Nomad clusters
• Fix permissions to use octet value when using data_with_permissions

version v0.5.41

• Add automatic push for build images to Kubernetes
• Add automatic refresh for Kubernetes jobs
• Add ability to expose local services to kubernetes clusters

version v0.5.40

• Update docs to implement new structure
• Add copy capability to build resources

version v0.5.39

• Update K3s to 1.27.4
• Update Nomad to 1.6.1

version v0.5.38

• Add new dev command to watch the given config for changes and automatically run up
• Change behavior of up to destroy any resources not in the current config

version v0.5.9

• Changed health checks to move checks like kubernetes pods to specific resources
• Added exec health check for containers

resource "container" {
  health_check {
    timeout = "30s"

    http {
      address       = "http://localhost:8500"
      success_codes = [200]
    }

    tcp {
      address = "localhost:8500"
    }

    exec {
      script = <<-EOF
        #!/bin/sh -e

        ls -las
      EOF
    }
  }
}

version v0.4.14

• Fix bug in latest Kubernetes release where DNS san for certificate did not contain the fully qualified name of the server

version v0.4.13

• Update Nomad version to 1.4.0
• Update K3s to v1.23.12
• Add capability to run K3s clusters when the Docker engine is not using overlay or overlay2

version v0.4.0

• Add new resource to enable copying resources

copy "testing" {
	source = "/"
	destination = "/path"

	permissions = "0700"
}

• Add new resource to enable generation of certificates

certificate_ca "testing" {
	output = "/"
}

certificate_leaf "testing" {
	ip_addresses = ["a","b"]
	dns_names = ["1","2"]

	ca_cert = "./file"
	ca_key = "./file"

	output = "/"
}

• Add new function to allow the creation of data directories with specific permissions

data_with_permissions("name","0777")

version v0.3.50

• Sanitize helm chart names replacing non acceptable characters

version v0.3.43

• Update Helm library to 3.8.2

version v0.3.42

• Add capability to install Helm charts from Helm repositories and local paths

helm "vault" {
  depends_on = ["helm.consul"] # only install one at a time

  cluster = "k8s_cluster.k3s"

  repository {
    name = "hashicorp"
    url  = "https://helm.releases.hashicorp.com"
  }

  chart   = "hashicorp/vault" # When repository specified this is the name of the chart
  version = "v0.18.0"         # Version of the chart when repository specified

  values = "./helm/vault-values.yaml"

  health_check {
    timeout = "240s"
    pods    = ["app.kubernetes.io/name=vault"]
  }
}

version v0.3.41

Shipyard now supports Podman through the Podman sock. Rootless and Root containers are both supported however, complex containers that require privileged access such as k8s_clusters and nomad_clusters required Root mode for podman.

version v0.3.40

• Do not attempt to import Docker images to Nomad and Kubernetes clusters when the Name is ""

version v0.3.37

• Adds capability for complex variable types and variable interpolation in templates. Previously, all vars in template resources were interpreted as strings. This change now allows richer types and where possible preserves the original type. For example, you can now define a variable with an array of integers and then loop over this array in a template.

variable "bool_var" {
	default = true
}

variable "num_var" {
	default = 13
}

variable "str_var" {
	default = "Abc"
}

variable "other_ports" {
	default = [2000,2001]
}

template "fetch_consul_resources" {
  source = <<EOF

bind_addr = "{{ GetPrivateInterfaces | attr \"address\" }}"
client_addr = "0.0.0.0"
data_dir = "#{{ .Vars.data_dir }}"
datacenter = "dc1"

enabled = #{{ .Vars.enabled }}
not_enabled = #{{ .Vars.not_enabled }}
bool_var = #{{ .Vars.bool_var }}
num_var = #{{ .Vars.num_var }}
string_var = "#{{ .Vars.string_var }}"

ports {
	#{{ range .Vars.ports }}
  grpc = #{{ . }}
	#{{ end }}
}

other_ports {
	#{{ range .Vars.other_ports }}
  grpc_other = #{{ . }}
	#{{ end }}
}
	EOF

	destination = "./out.txt"

  vars = {
		other_ports = var.other_ports
		bool_var = var.bool_var
		string_var = var.str_var
		num_var = var.num_var
		data_dir = "something"
		enabled = true
		not_enabled = false
		ports = [8502, 8500]
		port = 8342
		config = {
			a = 1
			names = ["foo","bar"]
		}
  }
}

* Adds new interpolation function `len`, this returns the length of an array or map variable.

* Upgrade default Kubernetes version to `v1.22.3`
* Upgrade default Nomad version to `v1.2.0`

version v0.3.35

• Fix bug with Nomad clusters when using exec or java drivers
• Update Nomad base version to 1.1.6
• Fix bug with ingress selecting ports for stopped jobs

version v0.3.34

• Fix bug where output variables were not honoring the disabled meta parameter

version v0.3.32

• Fix bug in Nomad job where status was incorrectly reported
• Improve error messages when Nomad Job validation fails
• Add support for Docker bind propagation

version v0.3.28

• Add bash completion for logs command
• Format status output to make clearer
• Add new Nomad version 1.1.3
• Return error message when resource has no name

version v0.3.1

• Set max random ports for clusters to 64000

version v0.3.1

Image Caching

To save bandwidth all containers launched from Kubernetes and Nomad clusters are cached by Shipyard. Currently images from the following registries are cached:

• k8s.gcr.io
• gcr.io
• asia.gcr.io
• eu.gcr.io
• us.gcr.io
• quay.io
• ghcr.io"
• docker.io

Previously only images created from the container resource were cached, this change will make substantial speed and bandwidth improvements when creating Shipyard resources once the image has been cached.

Shipyard launches a pull through cache when resources are created

Other changes

• Add max_restart_count to containers and sidecars
• Add function to retrieve the ipa
• Move terminal server to embedded in Shipyard binary
• Add capability to have a local terminal instance in Docs.

version 0.2.10

• Add capability to run local_exec commands as a daemon
• Add timeout to local_exec
• Move Creating TLS certificates message to debug log

version 0.2.0

• Add new interpolation function to get the ip address of the machine running shipyard
• Add CreateNamespace config parameter to Helm config

// ingress exposing a local application
// would enable traffic on the K8s cluster dc1 sent to:
//      my-local-service.shipyard.svc.cluster.local:9090
// to be directed to:
//      localhost:30002
// on the shipyard host
ingress "k8s-to-local" {
  source {
    driver = "k8s"
    
    config {
      cluster = "k8s_cluster.dc1"
      port = 9091
    }
  }
  
  destination {
    driver = "local"
    
    config { 
      address = "localhost"
      port = 30002
    }
  
  }
}

// ingress exposing an application on one K8s cluster to the shipyard host
// would enable traffic on the shipyard host sent to:
//      localhost:9090
// to be directed to:
//      dc1-service.mynamespace.svc.cluster.local:30002
// on the dc1 cluster
ingress "local-to-k8s" {
  source {
    driver = "local"
    
    config {
      port = 9092
    }
  }
  
  destination {
    driver = "k8s"
    
    config {
      cluster = "k8s_cluster.dc1"
      address = "k8s-to-local.shipyard.svc"
      port = 9091
    }
  }
}

• Add ability to define defaults for variables

Variables can also be defined in modules, variables specified as flags, files, or environment override defaults.

variable "mod_network" {
  default = "modulenetwork"
}

* New Interpolation functions file_path and file_dir

Ingress

Major refactor of ingress, implemented K8s and Local

version 0.1.18

• Abiility to run with external Docker engine. Shipyard now respects the environment variable DOCKER_HOST and can be run on external docker engines.

version 0.1.17

• Compress Shipyard binary to reduce distribution size
• Add ability to add variables overides with test command

version 0.1.15

Add updates for Nomad ingress

version 0.1.14

Various bug fixes

• Fix Nomad health checks
• Move version manager package

version 0.1.8

Fix local exec capability

Fixes bugs and extends the local exec capability for running commands local as part of a run

exec_remote "exec_standalone" {
  cmd = "consul"
  args = [
    "services",
    "register",
    "./config/redis.hcl"
  ]

  env {
    key = "CONSUL_HTTP_ADDR"
    value = "http://consul.container.shipyard.run:8500"
  }
}

version 0.0.32

Remote Exec

• Ensure remote exec containers use a unique ID

version 0.0.31

Containers

Add ability to configure network aliases for containers

  network { 
    name = "network.onprem"
    ip_address = "10.5.0.200"
    // Add network aliases for the container
    aliases = ["web.ingress.container.shipyard.run", "api.ingress.container.shipyard.run"]
  }

version 0.0.30

Bugfixes Exec

• Fix log output when executing commands in containers

version 0.0.28

Remote Exec

• Add working_directory parameter to allow setting of the execution folder when running commands

version 0.0.26

Push Command

• Changes to caching volumes inadvertently changed the behaviour of the push command. When an image was already cached in the cluster image volume push would not overwrite this image. If the user specifies the --force-update command then the pushed image would be overwritten regardles of local cache; however, force-update would also attempt to pull the image from a remote registry. Push was designed to push a local image to a cluster as part of the development flow, pulling remote images is not desireable behaviour.
• This change ensures that the push command will always push an image regardless of local cache.

version 0.0.25

Bugfixes

• Environment variables were not being correctly passed to remote container executions when attaching to an exising container

version 0.0.24

Env Command

Added a new command to simplify the export of environment variables. Environment variables can be defined in a blueprint readme, after the blueprint has completed the variables will be presented to the user.

Blueprint

---
title: Single Container Example
author: Nic Jackson
slug: container
browser_windows: http://consul-http.ingress.shipyard.run:8500
env:
  - KUBECONFIG=$HOME/.shipyard/config/k3s/kubeconfig.yaml
  - VAULT_ADDR=http://localhost:18200
  - VAULT_TOKEN=root
---

# Single Container

This blueprint shows how you can create a single container with Shipyard

Output

######################################################

Environment Variables

######################################################

This blueprint exports the following environment variables:

KUBECONFIG=$HOME/.shipyard/config/k3s/kubeconfig.yaml=
VAULT_ADDR=http://localhost:18200=
VAULT_TOKEN=root=

You can set exported environment variables for your current terminal session using the following command:

eval $(shipyard env)

The env command will write the defined environment variables to the command line...

➜ shipyard env
export KUBECONFIG=$HOME/.shipyard/config/k3s/kubeconfig.yaml
export VAULT_ADDR=http://localhost:18200
export VAULT_TOKEN=root

To set the variable to the command line the user can either copy and paste the output or evaulate it.

eval $(shipyard env)

Bugfixes

• Update terminal server in Docs, there was a bug with the latest Chrome where the terminal was not immediately displaying the terminal.

version 0.0.20

Clusters

• New feature to persist and share cache for importing images for clusters. Previously each cluster has its own cache which was created every time the cluster was started. Images were copied from the local machine to a Docker volume before being imported to the clsuter. If multiple clusters were defined in a config this process would run for each cluster unecessarilly slowed the startup time. Now images are cached in a persistant volume images.volume.shipyard.run improving startup times. When a cluster starts it first checks to see if an image exits, if it does not it will import the image from the local Docker instance.
• Added a new feature to the purge command which removes the persistent image cache.

Modules

• New resource type modules which can import other Shipyard configuration blueprints from a file or from GitHub.
• Modules do not respect dependencies, this feature is a TODO item.

module "k8s" {
	source = "github.com/jumppad-labs/jumppad/examples//single_k3s_cluster"
}

module "consul" {
	source = "../container"
}

version 0.0.19

Sidecar resource

Added a new resource sidecar. A sidecar is a special container which shares the network and ip address of the target container.

container "consul" {
  image   {
    name = "consul:1.6.1"
  }

  command = ["consul", "agent", "-config-file=/config/consul.hcl"]
}

sidecar "consul" {
  target = "container.consul"

  image   {
    name = "consul:1.6.1"
  }

  command = ["consul", "connect", "envoy", "-sidecar-for", "myservice"]
}

Helm charts

• Added ability to set the namespace

Purge Command

• Added shipyard purge command to remove all cached images, blueprints, and helm charts

Bugfixes

• Remote exec now correctly pulls a container

version 0.0.18

Helm charts

• Added capability to reference a github repo as well as local charts
• Added the ability to specify values as a map addition to a local file

helm "vault" {
  cluster = "k8s_cluster.k3s"
  chart = "github.com/hashicorp/vault-helm"

  values_string = {
    "server.dataStorage.size" = "128Mb",
    "server.dev.enabled" = "true",
    "server.standalone.enabled" = "true",
    "server.authDelegator.enabled" = "true"
  }

  health_check {
    timeout = "120s"
    pods = ["app.kubernetes.io/name=vault"]
  } 
}

General

• Added --force-update flag to run and get. The user can specify that any local cache for images and helm charts is ignored at runtime.

version 0.0.16

Bugfixes

• By default when looking up the id of a Docker container Docker does a greedy match, this caused issues where we would grab an incorrect id on destroy. Changed to use a regex.

version 0.0.16

Bugfixes

• Fixes a bug where local folders were being created for types other than bind
• Fix a bug where Volumes for Nomad Clusters were not changed to absolute paths

version 0.0.15

General

• When defining a volume mount for a container, if the source does not exist, creat it rather than erroring

version 0.0.14

Bugfixes

• Fix bug where Documentation resource was failing to create on OSX. Docker could not read temporary file. Moved temp files to .shipyard folder to avoid conflict.

version 0.0.11

Browser windows

Change OpenInBrowser to a string so that the user can specify the path

Version check

Check the latest verison on startup

########################################################
                   SHIPYARD UPDATE
########################################################

The current version of shipyard is "0.0.10", you have "841f3b0445cc01f44ea2728dd5113f6b0f611e1f".

To upgrade Shipyard please use your package manager or, 
see the documentation at:
https://shipyard.run/docs/install for other options.

version 0.0.9

Browser windows

Ensure that browser windows are openened correctly on Windows platforms.

Preflight

Check that Docker is installed and running

version 0.0.8

Bugfixes

• Fix problem with home folder resolution on Windows
• Add ipv6 DNS entries for resources

version 0.0.7

Ingress

Add three new ingress types nomad_ingress, k8s_ingress, container_ingress, these new type have cluster specific options rather than trying to shoehorn all into ingress. For now ingress exists and is backward compatible.

k8s_ingress "consul-http" {
  cluster = "k8s_cluster.k3s"
  service  = "consul-consul-server"

  network {
    name = "network.cloud"
  }

  port {
    local  = 8500
    remote = 8500
    host   = 18500
  }
}

container_ingress "consul-http" {
  target  = "container.consul"

  port {
    local  = 8500
    remote = 8500
    host   = 18500
  }

  network  {
    name = "network.cloud"
  }
}

nomad_ingress "nomad-http" {
  cluster  = "nomad_cluster.dev"
  job = ""
  group = ""
  task = ""

  port {
    local  = 4646
    remote = 4646
    host   = 14646
    open_in_browser = true
  }

  network  {
    name = "network.cloud"
  }
}

Browser Windows

• Added ability to open browser windows to ingress, containers, docs
• Only open browser windows if they have not been opened

General

• Can now do shipyard run or shipyard run . and this equals shipyard run ./
• Sidebar configuration is now optional in docs
• Changed names of Docker resources and FQDN for resources to [name].[type].shipyard.run
• Resources are now resolvable via DNS
• Added interpolation helper ${home()} to resolve the home folder from config
• Added interpolation helper ${shipyard()} to resolve the shipyard folder from config

Bug fixes

• Fix serialisation of blueprint in state
• Fix bug with rolled back resources being in an inconsistent state
• Fix bug with state not serialzing correctly

version 0.0.6

Bug fixes

• Fix bug where heirachy in state was lost on incremental applies
• Minor UX tweak for run command, use current directory as default

version 0.0.5

Container

HTTP health checks can now be defined for containers. At present HTTP health checks are executed on the client and require forwarded ports or an ingress to be configured. Future updates will move this functionality to the network removing the requirement for external access.

container "vault" {
  image {
    name = "hashicorp/vault-enterprise:1.4.0-rc1_ent"
  }

  command = [
    "vault",
    "server",
    "-dev",
    "-dev-root-token-id=root",
    "-dev-listen-address=0.0.0.0:8200",
  ]

  port {
    local = 8200
    remote = 8200
    host = 8200
  }

  health_check {
    timeout = "30s"
    http = "http://localhost:8200/v1/sys/health"
  }
}

Nomad Job

Nomad job is a new resource type which allows the running of Nomad Jobs on a cluster. Health checks can be applied to jobs a health check will only be marked as passed when all the tasks in side the job are reported as "Running".

Nomad job is a dependent resource and will not apply until the cluster is up and healthy.

nomad_cluster "dev" {
  version = "v0.10.2"

  nodes = 1 // default

  network {
    name = "network.cloud"
  }

  image {
    name = "consul:1.7.1"
  }
}

nomad_job "redis" {
  cluster = "nomad_cluster.dev"

  paths = ["./app_config/example2.nomad"]
  
  health_check {
    timeout = "60s"
    nomad_jobs = ["example_2"]
  }
}

version 0.0.4

Container

Allow HTTP health checks to be added to containers

version 0.0.3

This version was skipped due to issues getting Chocolately distributions setup

version 0.0.2

Docs

Improve UX with documentation, Shipyard now autogenerates the JSON files required for Docusarus, the user only needs to author the markdown

Nomad

Implmeneted ability to push a local container to the Nomad cluster Allow mounting of custom volumes for Nomad clusters

Build process

Added Chocolatey and Brew, Deb and Rpm instalation sources

Yard files

Yard files are to be depricated in favor of Markdown files for blueprints. The information which was previously added to the .yard file can now be added as frontdown in a README.md file which resides in the root of your blueprint.

---
title: Single Container Example
author: Nic Jackson
slug: container
browser_windows: http://localhost:8080
---

# Single Container

This blueprint shows how you can create a single container with Shipyard

```shell
curl localhost:8080
```

When the user runs shipyard run, this renders to the terminal as:

########################################################

Title Single Container Example
Author Nic Jackson

########################################################


1 Single Container
────────────────────────────────────────────────────────────────────────────────

This blueprint shows how you can create a single container with Shipyard

┃ curl localhost:8080

Bugfixes

• Move create shipyard home directory to run or get, this was generating with invalid permissions when using the quick install

version 0.0.0-beta.12

Bugfixes

• Ensure downloaded blueprints are stored with their full path
• Increase test coverage

version 0.0.0-beta.11

Bugfixes

• Correct log line in Kubernetes controller

version 0.0.0-beta.11

Bugfixes

• Fix bug where Kubernetes Config was not returning an error when applying bad config
• Add log output for Helm charts and Kubernetes config

version 0.0.0-beta.10

Added no-browser to run command

The run command now has a no-browser flag which supresses any browser windows from opening if they are defined in the stack

version 0.0.0-beta.9

Updated exec command

Exec command now uses lighter ingress container instead of tools

Container resource

Add "entrypoint" configuration to set the containers entrypoint

Docs

New documentation contianer which proxies terminal websockets using Envoy. This was an issue when the docs site was running behind a proxy server such as Instruqt.

version 0.0.0-beta.8

Updated status command

The status command now pretty prints the resources

➜ shipyard status

 [ CREATED ] docs.docs
 [ CREATED ] container.tools
 [ CREATED ] helm.consul
 [ CREATED ] ingress.consul-http
 [ CREATED ] k8s_cluster.k3s
 [ CREATED ] network.cloud
 [ CREATED ] container.vscode

Pending: 0 Created: 7 Failed: 0

To view status in json format use the --json flag

Bug fixes

• alpine/linux was pulled every time when importing images regardless of local cache
• fix push to use new configuration

version 0.0.0-beta.7

Helm provider

• Helm provider now uninstalls the chart when deleting a resource, previousuly it was assumed that a chart and cluster would be deleted together
• Added exec command to allow the creation of a shell or execution of a command in a container or pod

➜ yard-dev exec k8s_cluster.k3s consul-consul-227vz               
parameters: []string{"k8s_cluster.k3s", "consul-consul-227vz"} - command: []string{}
2020-02-19T11:45:28.523Z [DEBUG] Image exists in local cache: image=shipyardrun/tools:latest
2020-02-19T11:45:28.524Z [INFO]  Creating Container: ref=exec-524329800
2020-02-19T11:45:28.641Z [DEBUG] Attaching container to network: ref=exec-524329800 network=network.cloud
/ # ls -las
total 68
     4 drwxr-xr-x    1 root     root          4096 Feb 19 11:38 .
     4 drwxr-xr-x    1 root     root          4096 Feb 19 11:38 ..
     4 drwxr-xr-x    1 root     root          4096 Sep 13 06:21 bin

• Added version command to return the current application verion
• When restarting from pause, health check all containers, helm charts and k8s config
• Update status command to pretty print status and add --json flag for detail

Bug fixes

• Improve test quality

version 0.0.0-beta.6

Bug fixes

• Alpine container not pulled when copying images to cluster
• Health check for pod was only looking at status not ready checks
• Check Network exists before removing
• Upgrade Helm dependency

Version 0.0.0-beta.5

Introduce taint command and the ability to re-create resources.

Resources can now be tainted using the command shipyard taint [type] [name]

When a resource is marked as tained the next run of shipyard run will destroy the resource and re-create it. This feature is especailly useful when building blueprints, often you require a change to a particular container you run shipyard destroy to destroy the stack and then shipyard run to re-create. Now it is possible to destroy only the affected resource with shipyard taint.

Change behaviour when processing folders

Previously shipyard run would recurse into folders, this behaviour causes problems when the sub-folders contain *.hcl files which are not Shipyard resources. shipyard run now only process the top level folder. Sub folder support will be added when we add the module feature.

Improve handling for failed resources

Resources which fail to create can now be retired by re-running shipyard run, any depended resources which were not created due to the failure will also be created when the command is run.