看来，只有一个乘客实例服务请求，其他人只是服务502错误。 这会导致间歇性错误模式，因为只有针对第二个实例的请求失败。
~$ rvmsudo passenger-status ----------- General information ----------- max = 4 count = 2 active = 0 inactive = 2 Waiting on global queue: 0 ----------- Application groups ----------- /u/apps/pixie.strd6.com/current: App root: /u/apps/pixie.strd6.com/current * PID: 3179 Sessions: 0 Processed: 121 Uptime: 3m 57s * PID: 3762 Sessions: 0 Processed: 0 Uptime: 2s
更新到Rails 3.1.0 rc5后发生这种情况
2011/07/27 21:37:37 [error] 3125#0: *608 upstream prematurely closed connection while reading response header from upstream,client: 68.226.71.148,server: pixieengine.com,request: "GET /chats/recent HTTP/1.1",upstream: "passenger:unix:/passenger_helper_server:",host: "pixieengine.com",referrer: "http://pixieengine.com/projects/426/ide" 2011/07/27 21:38:31 [error] 3125#0: *596 upstream prematurely closed connection while reading response header from upstream,client: 76.102.14.57,referrer: "http://pixieengine.com/pixel-editor" 2011/07/27 21:39:12 [error] 3125#0: *576 upstream prematurely closed connection while reading response header from upstream,client: 68.8.173.234,referrer: "http://pixieengine.com/community/forums/1" 2011/07/27 21:39:12 [error] 3125#0: *687 upstream prematurely closed connection while reading response header from upstream,client: 201.231.103.247,request: "GET /chats/active_users HTTP/1.1",referrer: "http://pixieengine.com/projects/demo/ide" 2011/07/27 21:39:12 [error] 3125#0: *686 upstream prematurely closed connection while reading response header from upstream,referrer: "http://pixieengine.com/projects/demo/ide"
我做错了，还是在networking/ HTTP的错误？
可用的CSS，但不推出时推送到生产
使用Nginx工作在Phusion Passenger上的Sequel DB连接的正确位置
如何使用乘客在Nginx服务器上部署Rails项目？
标题variables在生产中丢失
一个用于多个域名的rails应用程序
Gitlab：运行Unicorn的问题，Resque with Passenger / Nginx
无法派生应用程序
Passenger + Nginx + Node（Meteor）100％的cpu使用率
我通过切换到Unicorn解决了这个问题。
我实际上并没有弄清楚如何修复乘客，但是我能够验证它是因为从Rails 3.0.9过渡到3.1.0.rc5而打破的。