Atlassian 在 AWS 上的解决方案

在 Atlassian，增长步入快车道。该公司每天都在增加更多的客户，因此需要一种简单的方法来扩展 JIRA，JIRA 每月以 15,000 个支持请求的速度增长。支持此站点的实例以前托管在数据中心中，这给扩展带来了挑战。 Atlassian 的技术客户经理 Brad Bressler 说：“我们增长的规模使得快速向应用程序添加节点变得很困难。” “这是我们面向客户的实例，它收集了我们全球产品的所有支持票。它是世界上最大的 JIRA 实例之一，在本地扩展和维护它变得越来越难。”例如，support.atlassian.com 实例托管在单个本地服务器上，公司需要经常关闭该服务器进行维护。

该公司还需要确保 JIRA 的高可用性。 Atlassian 的首席解决方案工程师 Neal Riley 说：“这是一个任务关键型应用程序，可能受到停机影响的客户数量巨大。” “随着我们的成长，我们越来越关注数据中心的弹性和灾难恢复能力。”

为了迁移到更具可扩展性、高可用性的环境，Atlassian 创建了 JIRA 数据中心，这是该应用程序的新企业版本。但是，JIRA 数据中心需要共享存储。 “我们需要一个共享文件系统，这样各个应用程序节点就可以拥有一个共享的配置文件信息、插件和附件的真实来源，”Riley 说。

Atlassian 还需要响应希望在 Amazon Web Services (AWS) 云上运行 JIRA 的客户。 “我们最初考察了几家供应商，但 AWS 显然是领导者，”Bressler 说。 “我们需要自动扩展和可靠性，而 AWS 为我们提供了这一点。”该公司将 JIRA 数据中心迁移到 AWS 云，在 Amazon Elastic Compute Cloud (Amazon EC2) 实例上运行所有应用程序节点。 Atlassian 利用 Auto Scaling 组来实现两个应用程序的自动扩展，并使用 Elastic Load Balancing 将应用程序流量重定向到 Amazon EC2 实例以获得一致的性能。

• Prior to using Amazon EFS as a shared file system for JIRA Data Center, Atlassian tested the solution internally. During testing, the company discovered the technology was simple to set up and enabled consistent throughput and capacity that stayed within threshold. “Once we went live, everything worked exactly as we expected it to,” says Bressler. “It was performant, resilient, and easy to set up, and it is easy to maintain.”

• Because it can more easily manage its JIRA instances in the cloud, Atlassian is putting more effort into enhancing applications. “By moving to the AWS Cloud, our company has been able to focus more on what we do well: providing great services to our customers,” says Bressler. “Instead of having to spend time on managing the back-end application stack, we can really step up our game and better support our tens of thousands of global customers.”

• Atlassian is better supporting its customers by utilizing the built-in disaster recovery and high availability of AWS. “We have better disaster-recovery capabilities and better uptime because our application data is replicated across multiple AWS Availability Zones,” says Riley. “If our application instances go down, we’re stopping thousands of people from getting support. By moving to a highly available platform on AWS, we are much more confident that our solutions are available at all times.”