7 commits

Author	SHA1	Message	Date
Adam Moussa	2cc849024f	fix(terraform): accept dumps that already contain data/forgejo.db ... Today's archive has no gitea-db.sqlite3 at the root. Copy that file only when it is present.	2026-09-29 19:37:48 -04:00
Adam Moussa	30f4b54052	fix(terraform): scope plan object reads and restore on the data volume ... The plan role only needs object reads for the verification zip. Unpacking a dump in /tmp fills the root volume.	2026-09-29 19:35:07 -04:00
Adam Moussa	8eb19b4e55	fix(terraform): let the plan role read object tags and retention ... The S3 provider refreshes tagging, ACL, attributes, and Object Lock on the Lambda zip.	2026-09-29 19:28:05 -04:00
Adam Moussa	566a669c7b	fix(terraform): let the plan role read backup object metadata ... HeadObject on the Lambda zip is s3:GetObject. The plan role only had the bucket ARNs.	2026-09-29 19:25:14 -04:00
Adam Moussa	d511861296	fix(terraform): let the plan role read bucket website config ... The S3 provider refreshes GetBucketWebsite. The plan role was denied on the three Forgejo buckets.	2026-09-29 19:22:34 -04:00
Adam Moussa	8360cfb1db	fix(terraform): allow HCP refresh of the log group and parameter ... The scoped apply role can create those resources, but CloudWatch and SSM list them on a wildcard ARN. DescribeLogGroups and DescribeParameters need that resource.	2026-09-29 19:10:08 -04:00
Adam Moussa	e4982b87ad	feat(terraform): migrate forgejo to HCP Terraform (PLAT-80) (#100) ... * feat(terraform): migrate forgejo to HCP Terraform Move the prod host onto workspace forgejo-prod in the After Hours VPC and freeze CDK push deploys so cutover can happen without applying into seahaven-prod. * fix(terraform): restore forgejo.db from the dump tarball Boot restore was copying data/ and repos/ and leaving the sqlite file at the archive root, so a volume restore started Forgejo with no database.	2026-09-29 22:49:21 +00:00