data pipeline refactor and fix

This commit is contained in:
2026-04-13 18:30:04 -04:00
parent 6418729b16
commit 326bf80846
96 changed files with 7107 additions and 1763 deletions

158
bin/deploy-all Executable file
View File

@@ -0,0 +1,158 @@
#!/usr/bin/env bash
set -e
SCRIPT_DIR="$(cd "$(dirname "${BASH_SOURCE[0]}")" && pwd)"
ROOT_DIR="$(cd "$SCRIPT_DIR/.." && pwd)"
# Colors
RED='\033[0;31m'
GREEN='\033[0;32m'
YELLOW='\033[1;33m'
BLUE='\033[0;34m'
NC='\033[0m' # No Color
KUBECTL="kubectl --context=prod"
CLEAR_SANDBOXES=0
usage() {
echo "Usage: $0 [--sandboxes]"
echo ""
echo "Deploy all services to production. Does NOT update secrets (use bin/secret-update)."
echo ""
echo "Steps performed:"
echo " 1. Apply base kustomize manifests (namespaces, RBAC, policies)"
echo " 2. Apply infrastructure.yaml (statefulsets, deployments)"
echo " 3. Run bin/config-update prod"
echo " 4. Build and deploy all application images"
echo " 5. Wait for rollouts"
echo ""
echo "Options:"
echo " --sandboxes Delete sandbox Deployments and Services (PVCs are retained)."
echo " The gateway will recreate sandboxes on next user login."
echo ""
exit 1
}
for arg in "$@"; do
case "$arg" in
--sandboxes)
CLEAR_SANDBOXES=1
;;
--help|-h)
usage
;;
*)
echo -e "${RED}Unknown argument: $arg${NC}"
usage
;;
esac
done
echo -e "${YELLOW}╔══════════════════════════════════════════╗${NC}"
echo -e "${YELLOW}║ PRODUCTION FULL DEPLOY ║${NC}"
echo -e "${YELLOW}╚══════════════════════════════════════════╝${NC}"
echo ""
echo -e "${YELLOW}⚠️ This will update ALL production services.${NC}"
echo -e "${YELLOW} Secrets are NOT updated (run bin/secret-update prod separately).${NC}"
if [ "$CLEAR_SANDBOXES" == "1" ]; then
echo -e "${YELLOW} Sandbox deployments will be DELETED (PVCs retained).${NC}"
fi
echo ""
read -p "Are you sure you want to continue? (yes/no): " confirm
if [[ "$confirm" != "yes" ]]; then
echo "Aborted."
exit 0
fi
step() {
echo ""
echo -e "${BLUE}━━━ $1 ━━━${NC}"
}
ok() {
echo -e "${GREEN}✓${NC} $1"
}
fail() {
echo -e "${RED}✗ $1${NC}"
exit 1
}
# ── Step 1: Base kustomize manifests ─────────────────────────────────────────
step "Step 1/5: Applying base kustomize manifests"
cd "$ROOT_DIR"
$KUBECTL apply -k deploy/k8s/prod/
ok "Base manifests applied (namespaces, RBAC, policies, quotas)"
# ── Step 2: Infrastructure ────────────────────────────────────────────────────
step "Step 2/5: Applying infrastructure.yaml"
$KUBECTL -n ai apply -f deploy/k8s/prod/infrastructure.yaml
ok "Infrastructure applied"
# ── Step 3: Configs ───────────────────────────────────────────────────────────
step "Step 3/5: Updating configs"
# config-update prod will prompt for confirmation; we already confirmed above,
# so feed "yes" automatically via stdin.
echo "yes" | "$SCRIPT_DIR/config-update" prod
ok "Configs updated"
# ── Step 4: Build and deploy all application images ───────────────────────────
step "Step 4/5: Building and deploying application images"
echo ""
SERVICES=(gateway web sandbox lifecycle-sidecar flink relay ingestor)
for service in "${SERVICES[@]}"; do
echo -e "${GREEN}→${NC} Deploying $service..."
"$SCRIPT_DIR/deploy" "$service" prod
ok "$service deployed"
echo ""
done
# ── Step 4b: Optionally clear sandbox deployments ─────────────────────────────
if [ "$CLEAR_SANDBOXES" == "1" ]; then
step "Step 4b: Clearing sandbox deployments"
SANDBOX_DEPLOYS=$($KUBECTL -n sandbox get deployments -o name 2>/dev/null || true)
SANDBOX_SVCS=$($KUBECTL -n sandbox get services -o name 2>/dev/null || true)
if [ -z "$SANDBOX_DEPLOYS" ]; then
echo " No sandbox deployments found."
else
echo " Deleting sandbox deployments..."
echo "$SANDBOX_DEPLOYS" | xargs $KUBECTL -n sandbox delete
ok "Sandbox deployments deleted"
fi
if [ -n "$SANDBOX_SVCS" ]; then
echo " Deleting sandbox services..."
echo "$SANDBOX_SVCS" | xargs $KUBECTL -n sandbox delete
ok "Sandbox services deleted"
fi
echo -e "${YELLOW} PVCs retained — gateway will recreate sandboxes on next login.${NC}"
fi
# ── Step 5: Wait for rollouts ─────────────────────────────────────────────────
step "Step 5/5: Waiting for rollouts"
ROLLOUTS=(
"deployment/gateway"
"deployment/ai-web"
"deployment/relay"
"deployment/ingestor"
"deployment/flink-jobmanager"
"deployment/flink-taskmanager"
)
for r in "${ROLLOUTS[@]}"; do
echo -e "${GREEN}→${NC} Waiting for $r..."
$KUBECTL -n ai rollout status "$r" --timeout=180s || echo -e "${YELLOW} ⚠ $r did not become ready within 3 minutes${NC}"
done
echo ""
echo -e "${GREEN}╔══════════════════════════════════════════╗${NC}"
echo -e "${GREEN}║ Deploy complete! ║${NC}"
echo -e "${GREEN}╚══════════════════════════════════════════╝${NC}"
echo ""
echo " Verify: curl -I https://dexorder.ai/api/health"
echo ""