All notable changes to this project will be documented in this file.
-
The format is based on Keep a Changelog.
-
This project adheres to Semantic Versioning.
- Support for multi-region, multi-zone deployments in GCP.
- Support for multi-zone deployments in AWS. Since multi-region deployments require separate provider blocks, we don't allow multiple regions to avoid increased repo complexity.
- Support for GPUs on GCP (via the terraform
accelerator
block) and AWS. Includes driver installation script and a gpu-specificdocker-compose.yaml
file to expose GPUs to the node container for diagnostics. - Terraform formatter in pipeline and README.
- Format of node specification in
.tfvars
. Nodes are now specified via a map (seevariables.tf
) where keys correspond to node IDs. - Format of router specification in
.tfvars
. Router is now specified via a map (seevariables.tf
). - Naming conventions for configuration
.json
files. One file per deployed node, names (without.json
postfix) matching the node IDs (keys ofnodes
fromvariables.tf
), are now the only requirements.
- All created resources are now parametrized by cluster name, so no conflicts arise from successive deployments within the same project.
- Omissions in Makefile.
- Initial release of Infernet Deploy.