There are two settings for scaling on Cycle:
- Through the container config.
- Through infrastructure auto-scale groups.
To dive deeper on either topic please see the links above, but from a high level these infrastructure auto-scale groups provide the mechanism that will add (and remove) additional compute resources to the cluster.
Auto-Scale Dashboard
To manage auto-scale groups, head over to the auto-scale dashboard.
To get there:
- Select the Infrastructure tab from the main, left-hand navigation.
- Select the Auto-Scaling tab from the infrastructure sub-navigation.
Groups Summary
At the top of the dashboard, users will find a list of auto-scaling groups for the hub for quick and easy navigation.
Next Step
Quick options to open the auto-scaling dropdown or create a new group.
Documentation
Curated links to documentation specifically scoped to subject matter for auto-scaling.
Create Auto-Scaling Group
Click the Add button to open the Create Auto-Scaling Group modal. From here a user can add the settings most appropriate for their group.
Field | Description |
|---|---|
Name | The name of the auto-scaling group. |
Identifier | A unique resource identifier for the auto-scaling group, important for CLI operations. |
Cluster | The infrastructure cluster with which the auto-scaling group is associated. |
Scale Up | The settings and conditions for adding more servers to the auto-scaling group. |
Scale Down | The settings and conditions for reducing the number of servers and managing server TTLs. |
Deployable Infrastructure
Users can add server models to the deployable infrastructure page. Each entry specifies a provider, a server model, one or more locations (with their availability zones), and a priority that determines the order Cycle attempts to deploy from when a scaling event requires new infrastructure. The list of deployable infrastructure is what Cycle uses when a scaling event comes in that requires infrastructure to be added. Servers added through auto-scaling also inherit features and constraints settings shown below.
The Add Model button will open a wizard, from which users can add their choices. Entries on this page can be edited after they are created.
Settings
Name and Identifier can be changed in the basic settings, but cluster cannot. For users who have not configured the Scale Up or Scale Down settings during create, those can be toggled and configured here as well.
The Scale Up settings for an auto-scaling group set the upper limit of how many servers can be created to support scaling demands for containers in the group.
The Scale Down settings expose a few more options.
Field | Description |
|---|---|
Min TTL | The minimum amount of time a server must remain online before it can be removed. |
Inactivity Period | The period of time a server can remain idle (without activity) before it is removed. |
Scale Down Method | The strategy used for scaling down servers, such as Last-In-First-Out (LIFO), First-In-First-Out (FIFO), or Default. |
Constraints
Constraints govern what is allowed to schedule onto the servers this auto-scale group provisions.
Setting | Description | Default / Notes |
|---|---|---|
Node tags applied to every server this group provisions. A container's deployment constraints ( | None selected | |
Whether these servers join the general (untagged) pool and can run workloads that don't explicitly target them. When off, only workloads that match this group's tags land here. | Off | |
Whether environment service containers (Load Balancer, Discovery, VPN, Scheduler, Gateway) may run on these servers. | Off | |
Permit oversubscribing resources scheduling containers whose reserves total more than the server physically has. Enables the Overcommit Multiple field. | Off |
Features
Features enable capabilities on the servers this group provisions.
Setting | Description | Default / Notes |
|---|---|---|
Allow the server to accept incoming SFTP connections (through the remote-volume integration). | Off | |
Override the size of the base volume — where Cycle stores container images on each server. Leave blank to use the platform default. | Blank = platform default which ends up being around 28GB usable space. | |
The gateway service mode for these servers. | Platform default is static and BGP is not yet supported. |