feat(gitops): implement NATS JetStream and TrueNAS staging provisioning (Phase 5)

This commit is contained in:
Tibo De Peuter 2026-08-05 18:13:15 +02:00
parent 2b533f4375
commit 2340c5a0e7
Signed by: tdpeuter
SSH key fingerprint: SHA256:u/h/LVoqKF1Iz02uOyxe6hcjmoZASCGV2HM0TG9ZMoU
3 changed files with 160 additions and 0 deletions

View file

@ -10,6 +10,36 @@ let
runtimeInputs = with pkgs; [ git opentofu coreutils ];
text = builtins.readFile ../../../../scripts/hypervisor-sync.sh;
};
stagingSyncScript = pkgs.writeShellApplication {
name = "staging-sync";
runtimeInputs = with pkgs; [ opentofu coreutils jq curl ];
text = builtins.readFile ../../../../scripts/staging-sync.sh;
};
natsConsumerScript = pkgs.writeShellApplication {
name = "nats-consumer";
runtimeInputs = with pkgs; [ natscli jq stagingSyncScript ];
text = ''
set -euo pipefail
NATS_URL=''${NATS_URL:-"nats://192.168.0.20:4222"}
echo "Starting NATS JetStream consumer for staging env..."
# Try to create stream and consumer if they don't exist
nats --server "$NATS_URL" stream add FORGEJO_EVENTS --subjects "forgejo.staging" --ack --max-msgs=-1 --max-bytes=-1 --max-age=1y --storage file -f || true
nats --server "$NATS_URL" consumer add FORGEJO_EVENTS STAGING --pull --ack explicit --filter forgejo.staging --deliver all -f || true
echo "Listening for messages..."
while true; do
# We use a simple sub to pull messages. In a real environment,
# a dedicated Go/Python client is better for manual explicit acks.
# This will auto-ack upon receipt and pass to the staging script.
nats --server "$NATS_URL" sub "forgejo.staging" | awk '/\[#.*\]/{flag=1; next} flag' | staging-sync || true
done
'';
};
in {
options.homelab.services.hypervisor-gitops = {
enable = mkEnableOption "Hypervisor GitOps Service";
@ -31,6 +61,8 @@ in {
environment.systemPackages = with pkgs; [
git
opentofu
natscli
jq
];
systemd.services.hypervisor-gitops = {
@ -60,5 +92,27 @@ in {
Persistent = true;
};
};
systemd.services.staging-sync = {
description = "Staging Environment NATS Consumer";
after = [ "network-online.target" ];
wants = [ "network-online.target" ];
wantedBy = [ "multi-user.target" ];
serviceConfig = {
Type = "simple";
User = "root"; # Needs root to read SOPS secrets
Restart = "always";
RestartSec = "10s";
StateDirectory = "hypervisor-gitops";
WorkingDirectory = "/var/lib/hypervisor-gitops";
# We assume TRUENAS_API_KEY is provided via a sops EnvironmentFile
# EnvironmentFile = config.sops.secrets."truenas-api-key".path;
ExecStart = "${natsConsumerScript}/bin/nats-consumer";
};
};
};
}

View file

@ -4,6 +4,10 @@ terraform {
source = "bpg/proxmox"
version = "~> 0.61.0"
}
truenas = {
source = "deevus/truenas"
version = "~> 0.1.0"
}
}
}
@ -17,10 +21,25 @@ variable "pr_number" {
type = string
}
variable "truenas_snapshot_id" {
description = "The ID of the TrueNAS snapshot to clone"
type = string
}
provider "proxmox" {
# Relies on PROXMOX_VE_ENDPOINT and PROXMOX_VE_API_TOKEN environment variables
}
provider "truenas" {
# Relies on TRUENAS_API_KEY and TRUENAS_BASE_URL environment variables
}
resource "truenas_dataset" "staging_clone" {
pool = "tank"
path = "production/staging-pr-${var.pr_number}"
snapshot_id = var.truenas_snapshot_id
}
resource "proxmox_virtual_environment_vm" "staging_vm" {
name = "staging-pr-${var.pr_number}"
description = "Ephemeral staging environment for PR #${var.pr_number}"

87
scripts/staging-sync.sh Normal file
View file

@ -0,0 +1,87 @@
#!/usr/bin/env bash
# This script is triggered by the NATS JetStream consumer on the Control Center.
# It reads a Forgejo webhook JSON payload from STDIN and orchestrates the
# Staging VM OpenTofu lifecycle.
set -euo pipefail
# Read JSON payload from STDIN
PAYLOAD=$(cat)
# Extract fields using jq
ACTION=$(echo "$PAYLOAD" | jq -r '.action // empty')
PR_NUMBER=$(echo "$PAYLOAD" | jq -r '.pull_request.number // empty')
if [ -z "$ACTION" ] || [ -z "$PR_NUMBER" ] || [ "$PR_NUMBER" == "null" ]; then
printf "Invalid or missing action/pr_number in payload. Exiting.\n"
exit 0
fi
# We only care about PR events
if [[ "$ACTION" != "opened" && "$ACTION" != "reopened" && "$ACTION" != "synchronized" && "$ACTION" != "closed" ]]; then
printf "Ignoring PR action: %%s\n" "$ACTION"
exit 0
fi
# Ensure workspace directory exists for this PR
WORKSPACE="/var/lib/hypervisor-gitops/staging-pr-${PR_NUMBER}"
OPENTOFU_SRC="/var/lib/hypervisor-gitops/nix-config/opentofu/staging-env"
# TrueNAS API variables
TRUENAS_URL=${TRUENAS_URL:-"https://192.168.0.11"}
TRUENAS_API_KEY=${TRUENAS_API_KEY:-""}
ZFS_DATASET=${ZFS_DATASET:-"tank/production"}
get_latest_snapshot() {
# Fetch the latest snapshot for the dataset from TrueNAS API
# Assumes TRUENAS_API_KEY is exported in the environment by SOPS
curl -sS -k -X GET \
-H "Authorization: Bearer ${TRUENAS_API_KEY}" \
-H "Accept: application/json" \
"${TRUENAS_URL}/api/v2.0/zfs/snapshot?id~=${ZFS_DATASET}%%25&limit=1&sort=-creation" | jq -r '.[0].id'
}
printf "Processing PR #%%s (Action: %%s)\n" "$PR_NUMBER" "$ACTION"
if [[ "$ACTION" == "closed" ]]; then
if [ ! -d "$WORKSPACE" ]; then
printf "Workspace %%s does not exist. Nothing to destroy.\n" "$WORKSPACE"
exit 0
fi
printf "Destroying Staging Environment for PR #%%s...\n" "$PR_NUMBER"
cd "$WORKSPACE"
tofu destroy -var="pr_number=${PR_NUMBER}" -auto-approve
# Cleanup
cd /
rm -rf "$WORKSPACE"
printf "Staging Environment Destroyed.\n"
exit 0
fi
# Provisioning (opened, reopened, synchronized)
printf "Setting up Staging Environment for PR #%%s...\n" "$PR_NUMBER"
if [ ! -d "$WORKSPACE" ]; then
mkdir -p "$WORKSPACE"
cp -r "$OPENTOFU_SRC"/* "$WORKSPACE"/
fi
cd "$WORKSPACE"
# Fetch latest TrueNAS snapshot dynamically
LATEST_SNAPSHOT=$(get_latest_snapshot)
if [ -z "$LATEST_SNAPSHOT" ] || [ "$LATEST_SNAPSHOT" == "null" ]; then
printf "Failed to retrieve the latest TrueNAS snapshot. Aborting.\n"
exit 1
fi
printf "Latest Snapshot ID: %%s\n" "$LATEST_SNAPSHOT"
# Apply OpenTofu
tofu init -upgrade
tofu apply -var="pr_number=${PR_NUMBER}" -var="truenas_snapshot_id=${LATEST_SNAPSHOT}" -auto-approve
printf "Staging Environment Provisioned successfully.\n"