Files
git.stella-ops.org/.gitea/scripts/release/rollback.sh

227 lines
6.0 KiB
Bash

#!/usr/bin/env bash
set -euo pipefail
# Rollback Script
# Sprint: CI/CD Enhancement - Deployment Safety
#
# Purpose: Execute rollback to a previous version
# Usage:
# ./rollback.sh --environment <env> --version <ver> --services <json> --reason <text>
#
# Exit codes:
# 0 - Rollback successful
# 1 - General error
# 2 - Invalid arguments
# 3 - Deployment failed
# 4 - Health check failed
SCRIPT_DIR="$(cd "$(dirname "${BASH_SOURCE[0]}")" && pwd)"
REPO_ROOT="$(cd "${SCRIPT_DIR}/../../.." && pwd)"
# Colors for output
RED='\033[0;31m'
GREEN='\033[0;32m'
YELLOW='\033[1;33m'
BLUE='\033[0;34m'
NC='\033[0m'
log_info() {
echo -e "${GREEN}[INFO]${NC} $*"
}
log_warn() {
echo -e "${YELLOW}[WARN]${NC} $*"
}
log_error() {
echo -e "${RED}[ERROR]${NC} $*" >&2
}
log_step() {
echo -e "${BLUE}[STEP]${NC} $*"
}
usage() {
cat << EOF
Usage: $(basename "$0") [OPTIONS]
Execute rollback to a previous version.
Options:
--environment <env> Target environment (staging|production)
--version <version> Target version to rollback to
--services <json> JSON array of services to rollback
--reason <text> Reason for rollback
--dry-run Show what would be done without executing
--help, -h Show this help message
Examples:
$(basename "$0") --environment staging --version 1.2.3 --services '["scanner"]' --reason "Bug fix"
$(basename "$0") --environment production --version 1.2.0 --services '["authority","scanner"]' --reason "Hotfix rollback"
Exit codes:
0 Rollback successful
1 General error
2 Invalid arguments
3 Deployment failed
4 Health check failed
EOF
}
# Default values
ENVIRONMENT=""
VERSION=""
SERVICES=""
REASON=""
DRY_RUN=false
# Parse arguments
while [[ $# -gt 0 ]]; do
case "$1" in
--environment)
ENVIRONMENT="$2"
shift 2
;;
--version)
VERSION="$2"
shift 2
;;
--services)
SERVICES="$2"
shift 2
;;
--reason)
REASON="$2"
shift 2
;;
--dry-run)
DRY_RUN=true
shift
;;
--help|-h)
usage
exit 0
;;
*)
log_error "Unknown option: $1"
usage
exit 2
;;
esac
done
# Validate required arguments
if [[ -z "$ENVIRONMENT" ]] || [[ -z "$VERSION" ]] || [[ -z "$SERVICES" ]]; then
log_error "Missing required arguments"
usage
exit 2
fi
# Validate environment
if [[ "$ENVIRONMENT" != "staging" ]] && [[ "$ENVIRONMENT" != "production" ]]; then
log_error "Invalid environment: $ENVIRONMENT (must be staging or production)"
exit 2
fi
# Validate services JSON
if ! echo "$SERVICES" | jq empty 2>/dev/null; then
log_error "Invalid services JSON: $SERVICES"
exit 2
fi
log_info "Starting rollback process"
log_info " Environment: $ENVIRONMENT"
log_info " Version: $VERSION"
log_info " Services: $SERVICES"
log_info " Reason: $REASON"
log_info " Dry run: $DRY_RUN"
# Record start time
START_TIME=$(date +%s)
# Rollback each service
FAILED_SERVICES=()
SUCCESSFUL_SERVICES=()
echo "$SERVICES" | jq -r '.[]' | while read -r service; do
log_step "Rolling back $service to $VERSION..."
if [[ "$DRY_RUN" == "true" ]]; then
log_info " [DRY RUN] Would rollback $service"
continue
fi
# Determine deployment method
HELM_RELEASE="stellaops-${service}"
NAMESPACE="stellaops-${ENVIRONMENT}"
# Check if Helm release exists
if helm status "$HELM_RELEASE" -n "$NAMESPACE" >/dev/null 2>&1; then
log_info " Using Helm rollback for $service"
# Get revision for target version
REVISION=$(helm history "$HELM_RELEASE" -n "$NAMESPACE" --output json | \
jq -r --arg ver "$VERSION" '.[] | select(.app_version == $ver) | .revision' | tail -1)
if [[ -n "$REVISION" ]]; then
if helm rollback "$HELM_RELEASE" "$REVISION" -n "$NAMESPACE" --wait --timeout 5m; then
log_info " Successfully rolled back $service to revision $REVISION"
SUCCESSFUL_SERVICES+=("$service")
else
log_error " Failed to rollback $service"
FAILED_SERVICES+=("$service")
fi
else
log_warn " No Helm revision found for version $VERSION"
log_info " Attempting deployment with specific version..."
# Try to deploy specific version
IMAGE_TAG="${VERSION}"
VALUES_FILE="${REPO_ROOT}/devops/helm/values-${ENVIRONMENT}.yaml"
if helm upgrade "$HELM_RELEASE" "${REPO_ROOT}/devops/helm/stellaops" \
-n "$NAMESPACE" \
--set "services.${service}.image.tag=${IMAGE_TAG}" \
-f "$VALUES_FILE" \
--wait --timeout 5m 2>/dev/null; then
log_info " Deployed $service with version $VERSION"
SUCCESSFUL_SERVICES+=("$service")
else
log_error " Failed to deploy $service with version $VERSION"
FAILED_SERVICES+=("$service")
fi
fi
else
log_warn " No Helm release found for $service"
log_info " Attempting kubectl rollout undo..."
DEPLOYMENT="stellaops-${service}"
if kubectl rollout undo deployment/"$DEPLOYMENT" -n "$NAMESPACE" 2>/dev/null; then
log_info " Rolled back deployment $DEPLOYMENT"
SUCCESSFUL_SERVICES+=("$service")
else
log_error " Failed to rollback deployment $DEPLOYMENT"
FAILED_SERVICES+=("$service")
fi
fi
done
# Calculate duration
END_TIME=$(date +%s)
DURATION=$((END_TIME - START_TIME))
# Summary
echo ""
log_info "Rollback completed in ${DURATION}s"
log_info " Successful: ${#SUCCESSFUL_SERVICES[@]}"
log_info " Failed: ${#FAILED_SERVICES[@]}"
if [[ ${#FAILED_SERVICES[@]} -gt 0 ]]; then
log_error "Failed services: ${FAILED_SERVICES[*]}"
exit 3
fi
log_info "Rollback successful"
exit 0