The Kubernetes platform.
The Package manager.
The Open Service Broker.
An Application is deployed to Drycc using
git push or the
Drycc Workflow can deploy any application or service that can run inside a container. In order to be scaled horizontally, applications must follow the Twelve-Factor App methodology and store any application state in external backing services.
For example, if your application persists state to the local filesystem -- common with content management systems like
Wordpress and Drupal -- it cannot be scaled horizontally using
Fortunately, most modern applications feature a stateless application tier that can scale horizontally inside Drycc.
Before deploying an application, users must first authenticate against the Drycc Controller using the URL supplied by their Drycc administrator.
$ drycc login http://drycc.example.com Opening browser to http://drycc.example.com/v2/login/drycc/?key=4ccc81ee2dce4349ad5261ceffe72c71 Waiting for login... .o.Logged in as admin Configuration file written to /root/.drycc/client.json
Drycc Workflow supports three different ways of building applications:
Cloud Native Buildpacks are useful if you want to follow cnb's docs for building applications.
Learn how to deploy applications using Buildpacks.
Dockerfiles are a powerful way to define a portable execution environment built on a base OS of your choosing.
Learn how to deploy applications using Dockerfiles.
Deploying a Container image onto Drycc allows you to take a Container image from either a public or a private registry and copy it over bit-for-bit, ensuring that you are running the same image in development or in your CI pipeline as you are in production.
Learn how to deploy applications using Container images.
It is possible to configure a few of the globally tunable settings on per application basis using
|DRYCC_DISABLE_CACHE||if set, this will disable the [imagebuilder cache] (default: not set)|
|DRYCC_DEPLOY_BATCHES||the number of pods to bring up and take down sequentially during a scale (default: number of available nodes)|
|DRYCC_DEPLOY_TIMEOUT||deploy timeout in seconds per deploy batch (default: 120)|
|IMAGE_PULL_POLICY||the kubernetes [image pull policy][pull-policy] for application images (default: "IfNotPresent") (allowed values: "Always", "IfNotPresent")|
|KUBERNETES_DEPLOYMENTS_REVISION_HISTORY_LIMIT||how many revisions Kubernetes keeps around of a given Deployment (default: all revisions)|
|KUBERNETES_POD_TERMINATION_GRACE_PERIOD_SECONDS||how many seconds kubernetes waits for a pod to finish work after a SIGTERM before sending SIGKILL (default: 30)|
Deploy timeout in seconds - There are 2 deploy methods, Deployments (see below) and RC (versions prior to 2.4) and this setting affects those a bit differently.
Deployments behave a little bit differently from the RC based deployment strategy.
Kubernetes takes care of the entire deploy, doing rolling updates in the background. As a result, there is only an overall deployment timeout instead of a configurable per-batch timeout.
The base timeout is multiplied with
DRYCC_DEPLOY_BATCHES to create an overall timeout. This would be 240 (timeout) * 4 (batches) = 960 second overall timeout.
This deploy timeout defines how long to wait for each batch to complete in
The base timeout is extended as well with healthchecks using
readiness where the bigger of those two is applied.
Additionally the timeout system accounts for slow image pulls by adding an additional 10 minutes when it has seen an image pull take over 1 minute. This allows the timeout values to be reasonable without having to account for image pull slowness in the base deploy timeout.
The advantage of Deployments is that rolling-updates will happen server-side in Kubernetes instead of in Drycc Workflow Controller, along with a few other Pod management related functionality. This allows a deploy to continue even when the CLI connection is interrupted.
Behind the scenes your application deploy will be built up of a Deployment object per process type, each having multiple ReplicaSets (one per release) which in turn manage the Pods running your application.
Drycc Workflow will behave the same way with
DRYCC_KUBERNETES_DEPLOYMENTS enabled or disabled (only applicable to versions prior to 2.4).
The changes are behind the scenes. Where you will see differences while using the CLI is
drycc ps:list will output Pod names differently.