Coder v1.21 (and later) feature satellites, which work in tandem with Networking v2, to provide low latency experiences for geo-distributed teams.
This article outlines the steps required to migrate from an
workspace provider deployment (used in Coder v1.20.x and earlier) to one using
the Networking v2 architecture present in Coder 1.21.x and later.
For in-depth information on Coder's networking changes (which appear in v1.21.0), see Rearchitecting Coder’s networking with WebRTC.
Each region that currently has an
envproxy-based workspace provider deployment
must be replaced with a satellite deployment (the satellite deployment is tasked
with the same proxy-related tasks for the workspaces located in that region).
Satellites are only compatible with workspace providers with Networking v2 enabled.
Developers will always have the lowest latency possible by connecting to the
Coder deployment closed to them geographically. Therefore, they should use the
access URL for the satellite deployment (e.g., those in the
australia-southeast1 region would use
coder-sydney.example.com) instead of
the access URL for the primary Coder deployment (e.g.,
Install the following dependencies if you haven't already:
To migrate, you must:
Though we expect you to see no downtime, end-users may experience negative impact to latency during the migration process for their region.
The first step is to deploy new satellites into each region to which you're expanding. To do so, follow the steps outlined in satellite management.
We recommend deploying the satellite into a namespace that's different from the one you're using for the existing workspace provider. Furthermore, because the satellite doesn't have any dependencies on the workspaces, you can deploy a satellite to any cluster and any namespace.
Log into Coder as a site manager, and go to Manage > Providers. Select the workspace provider, click the vertical ellipsis to its right, and select Edit. Enable the NetworkingV2 toggle and click Update Provider.
At this point, rebuild a workspace to ensure connectivity between the workspace provider and the workspace. Note that latency to the workspace may be negatively impacted until users connect to the new satellite deployments.
Repeat this step for each of your workspace providers.
Developers should now be able to connect to their nearest region using the
satellite's hostname (e.g.,
coder-sydney.example.com) and can expect low
latency when using their workspaces.
As of v1.21, Coder will no longer update the
coder/workspace-providerHelm chart. However, this Helm chart must stay deployed to allow for rollback opportunities. Coder will issue instructions on how to remove this chart beginning with v1.22.