@claude — FuzeFront node provisioning (fuzefront-worker-2) is fully working end-to-end EXCEPT the product tier: the handler's terraform apply reaches Contabo instance-create, but every product ID currently in config/infra-request-whitelist.json (V1, V45, V46, V47) returns Bad Request Product <X> is not available in region EU. The whitelist holds stale/unavailable product IDs.
Ask: add V92 and V76 to allowed_product_ids in config/infra-request-whitelist.json. Both are valid EU Cloud VPS products per Contabo's official Terraform provider examples; V92 ≈ 8 vCPU / 30 GB is the tier FuzeFront actually needs (the smaller whitelisted ones weren't strong enough anyway).
Before finalizing, confirm availability with cntb (availability is account/region-specific): e.g. cntb get products / the create-instance product list, since even V47 etc. were 'not available'. Then let FuzeFront know which tier(s) are actually orderable so node-requests.json can pin the right one. (Today FuzeFront's deploy/terraform/node-requests.json = V92.)
Note: a gated PR is likely already open from FuzeFront's latest dispatch (V92 isn't whitelisted yet → the handler routes it to a gated terraform plan for manual approval). Whitelisting V92 makes it auto-apply instead.
STATE (FuzeFront side, done): deploy/terraform now (1) wires the contabo-k3s-node module's required args (contabo creds + k3s join), (2) declares an S3 remote backend (backend "s3" {}) so node state is cloud-persisted via your handler's -backend-config (no more local/ephemeral CI state). Acceptance: V92 (+V76) in allowed_product_ids, cntb-confirmed available, fuzefront-worker-2 provisions + joins.
@claude — FuzeFront node provisioning (fuzefront-worker-2) is fully working end-to-end EXCEPT the product tier: the handler's terraform apply reaches Contabo instance-create, but every product ID currently in
config/infra-request-whitelist.json(V1, V45, V46, V47) returnsBad Request Product <X> is not availablein region EU. The whitelist holds stale/unavailable product IDs.Ask: add V92 and V76 to
allowed_product_idsinconfig/infra-request-whitelist.json. Both are valid EU Cloud VPS products per Contabo's official Terraform provider examples; V92 ≈ 8 vCPU / 30 GB is the tier FuzeFront actually needs (the smaller whitelisted ones weren't strong enough anyway).Before finalizing, confirm availability with
cntb(availability is account/region-specific): e.g.cntb get products/ the create-instance product list, since even V47 etc. were 'not available'. Then let FuzeFront know which tier(s) are actually orderable so node-requests.json can pin the right one. (Today FuzeFront's deploy/terraform/node-requests.json = V92.)Note: a gated PR is likely already open from FuzeFront's latest dispatch (V92 isn't whitelisted yet → the handler routes it to a gated
terraform planfor manual approval). Whitelisting V92 makes it auto-apply instead.STATE (FuzeFront side, done): deploy/terraform now (1) wires the contabo-k3s-node module's required args (contabo creds + k3s join), (2) declares an S3 remote backend (
backend "s3" {}) so node state is cloud-persisted via your handler's -backend-config (no more local/ephemeral CI state). Acceptance: V92 (+V76) in allowed_product_ids, cntb-confirmed available, fuzefront-worker-2 provisions + joins.