Glean does not currently support GovCloud regions.
Required AWS Services
When deciding on deployment region for your Glean deployment, you must ensure that all AWS services required by Glean are present.Service | Required |
---|---|
Amazon Athena | Yes |
Amazon Bedrock and Bedrock Runtime | No - Required only if Claude is the selected LLM |
AWS Certificate Manager (ACM) | Yes |
AWS CloudFormation | Yes |
AWS CloudTrail | No - Only used by debug roles for troubleshooting and can be blocked |
AWS CodeBuild | Yes |
DynamoDB | Yes (only used for Terraform state locking) |
Amazon Elastic Block Store (Amazon EBS) | Yes |
Amazon Elastic Compute Cloud (Amazon EC2) | Yes |
Amazon Elastic Container Registry (Amazon ECR and DKR) | Yes |
Amazon Elastic Kubernetes Service (Amazon EKS) | Yes |
Amazon ElastiCache | Yes |
Elastic Load Balancing/Application Load Balancing (ELB/ALB) | Yes |
Glue | Yes |
AWS Key Management Service (AWS KMS) | Yes |
AWS IAM | Yes |
AWS Lambda | Yes |
Amazon CloudWatch Metrics and Logs | Yes |
Amazon Relational Database Service (Amazon RDS & Performance Insights) | Yes |
AWS Resource Access Manager (RAM) | No - only if sharing TGW for on-prem connectivity |
Amazon Simple Storage Service (Amazon S3) | Yes |
Amazon SageMaker | Yes (see below for specifics) |
AWS Secrets Manager | Yes |
AWS Signer | Yes |
Amazon Simple Notification Service (Amazon SNS) | Yes |
Amazon Simple Queue Service (Amazon SQS) | Yes |
STS | Yes |
AWS Systems Manager (Session Manager only w/ SSM Messages and EC2 Messages) | Yes |
Amazon Textract | No |
Amazon Virtual Private Cloud (Amazon VPC) | Yes |
AWS WAFv2 | Yes |
Amazon Bedrock
Amazon Bedrock is only required if you wish to use Anthropic’s Claude models for your organization’s selected LLM. If so, you must ensure that the following models are present in Bedrock for your selected region (Glean does not support the region if these models are not available):- Claude 3.5 Sonnet (V1)
- Claude 3 Haiku
- Amazon Titan Embeddings G1 - Text
Amazon SageMaker
For Amazon Sagemaker, Glean requires that the following resource types be available in region:- ml.p3.16xlarge (NVIDIA V100 Multi GPU machines)
- ml.g4dn.xlarge
- ml.m5.2xlarge (deployments under 50M documents)
- ml.m5.4xlarge (deployments over 50M documents)