技术控

    今日:80| 主题:49270
收藏本版 (1)
最新软件应用技术尽在掌握

[其他] An Open Source, Self-Hosted Heroku

[复制链接]
梦三年忆三年 发表于 2016-10-14 03:44:43
134 3

立即注册CoLaBug.com会员,免费获得投稿人的专业资料,享用更多功能,玩转个人品牌!

您需要 登录 才可以下载或查看,没有帐号?立即注册

x
Running our own Heroku… It shouldn't be that hard, right?
  We have a small set of servers we use to run our internal applications. Nothing too complex, just monitoring, our ELK stack, Jenkins, and a few internal services.
  Given our rather modest requirements it may seem obvious that our first attempt at deployment automation, Chef, was a bit overkill for our needs. Not only that, we also wanted our engineers to be able to easily deploy applications to our servers without having to set up a Chef recipe — like the role Heroku plays in many of our client projects. We could have decided to run our internal applications on Heroku as well, but their pricing model wasn’t compatible with our relatively small-scale requirements.
  We researched a number of solutions and decided to go with Kubernetes. Having read positive reviews it initially seemed like a great solution to many of the problems we had encountered with Chef:
  
       
  • Everything ran in containers   
  • The system automatically managed multiple instances of individual applications   
  • The configuration was much less complex than Chef  
  However, a primary goal of our solution was to automate the process of going from code in a repository to a deployed application in our cluster — much like Heroku. Kubernetes did not appear have any mechanism to support this type of workflow, nor a handful of other features we had hoped to find.
  Not all who wander are lost, but we were...

   Clearly we were missing something, so we went to StackOverflow to search for answers .
   
An Open Source, Self-Hosted Heroku-1 (compatible,positive,problems,services,solution)

  As you can see from this incredibly helpful answer from Bill Prin on StackOverflow, what we actually wanted was a PaaS (Platform-as-a-Service), and Kubernetes only provided a piece of that functionality.
  That realization opened a whole new ecosystem of solutions to investigate. After researching several potential solutions, we tried a number of different options, but nothing seemed to fit our scale or use cases as well as we had hoped.
  We first attempted Openshift, which recommended a minimum configuration of far more and larger servers than were cost effective for the type of solution we wanted to create. Additionally our team wasn’t looking to navigate the many different versions of Openshift available: Openshift Enterprise, Openshift Origin, Openshift Online, and Openshift 2.
  We then looked at Dokku, but realized that only a single host was supported, and we definitely wanted the option of a cluster with multiple, redundant servers.
   Each solution had a drawback that prevented it from meeting all of our requirements, until we happened upon Flynn .
  It seemed like exactly what we were looking for: an open-source Heroku that we can run on our own infrastructure. Flynn even uses Heroku's buildpacks, which allow us to easily hand off our clients’ projects onto infrastructure they can manage themselves.
  Flynn was a bit young (pre-1.0 when we began), but it seemed to fit our needs perfectly. We installed it on a small server cluster and began deploying a few of our applications to the new Flynn PaaS. We had a few hiccups along the way, but with some amazing help from the team at Flynn, everything was up and running smoothly. With the release of 1.0 during our integration, its stability improved even further.
  Despite Flynn’s comprehensive featureset however, Chef was still required to automate certain tasks. Flynn does not support server monitoring or alerting, and there is no way to redirect Flynn's application logs to a service such as Elasticsearch.
  However, we are looking forward to a number of features on Flynn’s development roadmap, including support for persistent volumes and user accounts. Persistent volumes would allow us to run Elasticsearch within Flynn instead of managing it separately via Chef. User accounts will also be a huge step forward toward making Flynn a more mature and complete solution.
  While it may not meet each and every one of our requirements, and certainly hasn’t reached feature-parity with Heroku (yet), Flynn has performed very well for us thus far and will be an integral part of our production infrastructure for the foreseeable future.
   We encourage you to learn more at https://flynn.io/ orcontact us to learn how it could improve your architecutre.
友荐云推荐




上一篇:Static types in Python, oh my(py)
下一篇:VMware Horizon 7 Blast Extreme Primer—Everything an Admin Needs to Know
酷辣虫提示酷辣虫禁止发表任何与中华人民共和国法律有抵触的内容!所有内容由用户发布,并不代表酷辣虫的观点,酷辣虫无法对用户发布内容真实性提供任何的保证,请自行验证并承担风险与后果。如您有版权、违规等问题,请通过"联系我们"或"违规举报"告知我们处理。

如果声音不记得 发表于 2016-10-14 08:02:10
买买买
回复 支持 反对

使用道具 举报

幼枫 发表于 2016-10-27 03:13:33
支持支持再支持
回复 支持 反对

使用道具 举报

董习丽 发表于 2016-11-14 21:34:17
求加虫币!
回复 支持 反对

使用道具 举报

*滑动验证:
您需要登录后才可以回帖 登录 | 立即注册

本版积分规则

我要投稿

推荐阅读

扫码访问 @iTTTTT瑞翔 的微博
回页顶回复上一篇下一篇回列表手机版
手机版/CoLaBug.com ( 粤ICP备05003221号 | 文网文[2010]257号 )|网站地图 酷辣虫

© 2001-2016 Comsenz Inc. Design: Dean. DiscuzFans.

返回顶部 返回列表