hearth/scripts/deploy-all.sh

#!/usr/bin/env bash
#
# Full Infrastructure Deployment (RFC 0044)
#
# Deploys all phases in order with validation between each phase.
# This is the recommended way to perform a complete infrastructure rollout.
#
# Usage:
#   ./deploy-all.sh [--dry-run] [--skip-validation] [--start-phase <1-5>]
#
# Prerequisites:
#   - AWS CLI configured with appropriate credentials
#   - Terraform >= 1.6.0
#   - kubectl
#   - Helm 3.x
#
set -euo pipefail

SCRIPT_DIR="$(cd "$(dirname "${BASH_SOURCE[0]}")" && pwd)"

# Colors for output
RED='\033[0;31m'
GREEN='\033[0;32m'
YELLOW='\033[1;33m'
BLUE='\033[0;34m'
NC='\033[0m'

# Flags
DRY_RUN=false
SKIP_VALIDATION=false
START_PHASE=1

# Parse arguments
while [[ $# -gt 0 ]]; do
    case $1 in
        --dry-run)
            DRY_RUN=true
            shift
            ;;
        --skip-validation)
            SKIP_VALIDATION=true
            shift
            ;;
        --start-phase)
            START_PHASE="$2"
            shift 2
            ;;
        -h|--help)
            echo "Usage: $0 [--dry-run] [--skip-validation] [--start-phase <1-5>]"
            echo ""
            echo "Options:"
            echo "  --dry-run           Show what would be done without making changes"
            echo "  --skip-validation   Skip validation between phases (not recommended)"
            echo "  --start-phase <N>   Start from phase N (for resuming failed deployments)"
            exit 0
            ;;
        *)
            echo -e "${RED}Unknown option: $1${NC}"
            exit 1
            ;;
    esac
done

log_info() {
    echo -e "${BLUE}[INFO]${NC} $1"
}

log_success() {
    echo -e "${GREEN}[SUCCESS]${NC} $1"
}

log_warn() {
    echo -e "${YELLOW}[WARN]${NC} $1"
}

log_error() {
    echo -e "${RED}[ERROR]${NC} $1"
}

run_phase() {
    local phase="$1"
    local script="$2"
    local validation="$3"

    echo ""
    echo "========================================"
    echo "Starting Phase $phase"
    echo "========================================"
    echo ""

    # Run deployment
    local args=()
    if [ "$DRY_RUN" = true ]; then
        args+=("--dry-run")
    fi

    if [ -x "$script" ]; then
        "$script" "${args[@]}"
    else
        log_error "Deployment script not found or not executable: $script"
        exit 1
    fi

    # Run validation
    if [ "$SKIP_VALIDATION" = false ] && [ -x "$validation" ]; then
        echo ""
        log_info "Running validation for Phase $phase..."
        if ! "$validation"; then
            log_error "Phase $phase validation failed"
            echo ""
            echo "Options:"
            echo "  1. Fix the issues and re-run: $0 --start-phase $phase"
            echo "  2. Skip validation: $0 --skip-validation --start-phase $phase"
            echo "  3. Rollback: ./rollback.sh --phase $phase"
            exit 1
        fi
    fi

    # Tag successful deployment
    if [ "$DRY_RUN" = false ]; then
        log_info "Tagging Phase $phase deployment..."
        local tag_name="v0.${phase}.0-phase${phase}"
        local tag_message="Phase ${phase}: $(get_phase_name "$phase")"
        git tag -a "$tag_name" -m "$tag_message" 2>/dev/null || log_warn "Tag $tag_name already exists"
    fi

    log_success "Phase $phase complete"
}

get_phase_name() {
    case "$1" in
        1) echo "Foundation Infrastructure" ;;
        2) echo "Core Services" ;;
        3) echo "DNS and Email" ;;
        4) echo "Observability" ;;
        5) echo "E2EE Webmail" ;;
        *) echo "Unknown" ;;
    esac
}

main() {
    echo "========================================"
    echo "Full Infrastructure Deployment"
    echo "RFC 0044: Infrastructure Rollout Guide"
    echo "========================================"
    echo ""
    echo "This script will deploy all infrastructure phases in order:"
    echo "  Phase 1: Foundation Infrastructure (RFC 0039)"
    echo "  Phase 2: Core Services (RFC 0040)"
    echo "  Phase 3: DNS and Email (RFC 0041)"
    echo "  Phase 4: Observability (RFC 0042)"
    echo "  Phase 5: E2EE Webmail (RFC 0043) [optional]"
    echo ""

    if [ "$DRY_RUN" = true ]; then
        log_warn "Running in DRY-RUN mode - no changes will be made"
        echo ""
    fi

    if [ "$START_PHASE" -gt 1 ]; then
        log_info "Starting from Phase $START_PHASE"
        echo ""
    fi

    # Track start time
    local start_time
    start_time=$(date +%s)

    # Phase 1: Foundation
    if [ "$START_PHASE" -le 1 ]; then
        run_phase 1 "$SCRIPT_DIR/deploy-phase1-foundation.sh" "$SCRIPT_DIR/validate-phase1.sh"
    fi

    # Phase 2: Core Services
    if [ "$START_PHASE" -le 2 ]; then
        run_phase 2 "$SCRIPT_DIR/deploy-phase2-core-services.sh" "$SCRIPT_DIR/validate-phase2.sh"
    fi

    # Phase 3: DNS and Email
    if [ "$START_PHASE" -le 3 ]; then
        run_phase 3 "$SCRIPT_DIR/deploy-phase3-dns-email.sh" "$SCRIPT_DIR/validate-phase3.sh"
    fi

    # Phase 4: Observability
    if [ "$START_PHASE" -le 4 ]; then
        run_phase 4 "$SCRIPT_DIR/deploy-phase4-observability.sh" "$SCRIPT_DIR/validate-phase4.sh"
    fi

    # Phase 5: E2EE Webmail (optional)
    if [ "$START_PHASE" -le 5 ]; then
        echo ""
        log_info "Phase 5 (E2EE Webmail) is optional."
        read -p "Deploy Phase 5? (yes/no): " deploy_phase5
        if [ "$deploy_phase5" = "yes" ]; then
            run_phase 5 "$SCRIPT_DIR/deploy-phase5-e2ee-webmail.sh" "$SCRIPT_DIR/validate-phase5.sh"
        else
            log_info "Skipping Phase 5"
        fi
    fi

    # Calculate duration
    local end_time
    end_time=$(date +%s)
    local duration=$((end_time - start_time))
    local minutes=$((duration / 60))
    local seconds=$((duration % 60))

    # Final release tag
    if [ "$DRY_RUN" = false ]; then
        log_info "Creating final release tag..."
        git tag -a "v1.0.0" -m "Initial Release: Full Infrastructure Stack" 2>/dev/null || log_warn "Tag v1.0.0 already exists"
    fi

    echo ""
    echo "========================================"
    log_success "Full Infrastructure Deployment Complete!"
    echo "========================================"
    echo ""
    echo "Deployment Summary:"
    echo "  Duration: ${minutes}m ${seconds}s"
    echo "  Phases deployed: $(($START_PHASE > 1 ? 6 - $START_PHASE : 5))"
    echo ""
    echo "Post-Deployment Checklist:"
    echo "  [ ] All Prometheus targets UP"
    echo "  [ ] No firing alerts in Alertmanager"
    echo "  [ ] Grafana dashboards showing data"
    echo "  [ ] CockroachDB cluster healthy (3 nodes)"
    echo "  [ ] Vault unsealed and HA"
    echo "  [ ] Keycloak SSO working for all services"
    echo "  [ ] DNS resolving correctly"
    echo "  [ ] Email sending/receiving works"
    echo "  [ ] Backups running"
    echo ""
    echo "Documentation:"
    echo "  - Runbooks: $SCRIPT_DIR/../runbooks/"
    echo "  - RFC 0044: Infrastructure Rollout Guide"
}

main "$@"