Heroku路由器错误记录由LogEntries记录H13

Question

我想弄清楚我从LogEntries定期获得的这些错误值得担心。我有一个iPhone应用程序与Heroku上的Rails API进行对话。我使用HireFire来自动增加和减少应用程序需要的dynos和工作器的数量。

我认为这些erros是由dynos缩小，然后调用该dyno并因此导致H13错误引起的。如果是这样，是否有办法阻止它？用户的请求肯定不会不再存在？

这里是最常见的日志我在上下文中得到：

2012-10-18 10:01:15.170 209 <13>1 2012-10-18T10:01:14+00:00 app web.1 - - Started GET "/hirefireapp/xxxxxxxxxxxxxxxx/info" for 23.22.73.241 at Thu Oct 18 03:01:14 -0700 2012 
2012-10-18 10:01:15.170 162 <13>1 2012-10-18T10:01:14+00:00 app web.1 - - cache: [GET /hirefireapp/xxxxxxxxxxxxxxxx/info] miss 
2012-10-18 10:01:22.147 146 <13>1 2012-10-18T10:01:22+00:00 app web.2 - - Connected to NewRelic Service at collector-2.newrelic.com:80 
2012-10-18 10:01:22.308 217 <13>1 2012-10-18T10:01:22+00:00 app web.2 - - ** [NewRelic][10/18/12 03:01:22 -0700 xxxxxxxxxxxxxxxx (2)] INFO : Reporting performance data every 60 seconds. 
2012-10-18 10:01:38.118 86 <13>1 2012-10-18T10:01:38+00:00 app web.2 - - 
2012-10-18 10:01:38.212 86 <13>1 2012-10-18T10:01:38+00:00 app web.2 - - 
2012-10-18 10:01:38.212 154 <13>1 2012-10-18T10:01:38+00:00 app web.2 - - Started GET "/" for 204.93.223.151 at Thu Oct 18 03:01:38 -0700 2012 
2012-10-18 10:01:38.212 128 <13>1 2012-10-18T10:01:38+00:00 app web.2 - - Processing by CmsController#index as */* 
2012-10-18 10:01:38.212 148 <13>1 2012-10-18T10:01:38+00:00 app web.2 - - Rendered cms/index.html.erb within layouts/application (4.8ms) 
2012-10-18 10:01:41.291 124 <40>1 2012-10-18T10:01:41+00:00 heroku web.2 - - Stopping all processes with SIGTERM 
2012-10-18 10:01:42.243 217 <158>1 2012-10-18T10:01:42+00:00 heroku router - - Error H13 (Connection closed without response) -> GET my-api.heroku.com/ dyno=web.2 queue= wait= service= status=503 bytes= 

2.

2012-10-16 17:31:25.184 189 <13>1 2012-10-16T17:31:25+00:00 app web.2 - - ** [NewRelic][10/16/12 10:31:25 -0700 (2)] INFO : Dispatcher: thin 
2012-10-16 17:31:25.184 197 <13>1 2012-10-16T17:31:25+00:00 app web.2 - - ** [NewRelic][10/16/12 10:31:25 -0700 (2)] INFO : Application: my-api 
2012-10-16 17:31:25.184 220 <13>1 2012-10-16T17:31:25+00:00 app web.2 - - ** [NewRelic][10/16/12 10:31:25 -0700 (2)] INFO : New Relic Ruby Agent 3.3.1 Initialized: pid = 2 
2012-10-16 17:31:25.585 265 <13>1 2012-10-16T17:31:25+00:00 app web.2 - - ** [NewRelic][10/16/12 10:31:25 -0700 (2)] INFO : NewRelic::Agent::Samplers::DelayedJobLockSampler sampler not available: No DJ worker present 
2012-10-16 17:31:37.068 217 <13>1 2012-10-16T17:31:37+00:00 app web.2 - - ** [NewRelic][10/16/12 10:31:36 -0700 (2)] INFO : Reporting performance data every 60 seconds. 
2012-10-16 17:31:39.786 86 <13>1 2012-10-16T17:31:39+00:00 app web.2 - - 
2012-10-16 17:31:39.884 86 <13>1 2012-10-16T17:31:39+00:00 app web.2 - - 
2012-10-16 17:31:39.884 154 <13>1 2012-10-16T17:31:39+00:00 app web.2 - - Started GET "/" for xxx.xx.xxx.xxx at Tue Oct 16 10:31:39 -0700 2012 
2012-10-16 17:31:39.884 128 <13>1 2012-10-16T17:31:39+00:00 app web.2 - - Processing by CmsController#index as */* 
2012-10-16 17:31:39.981 149 <13>1 2012-10-16T17:31:39+00:00 app web.2 - - Rendered cms/index.html.erb within layouts/application (18.1ms) 
2012-10-16 17:31:46.082 217 <158>1 2012-10-16T17:31:46+00:00 heroku router - - Error H13 (Connection closed without response) -> GET my-api.heroku.com/ dyno=web.2 queue= wait= service= status=503 bytes= 

任何建议，将不胜感激！

感谢 皮特

的H13问题

Answer 1

的一个潜在原因是机架密钥空间。 Rack通过发送带有HTTP请求的大量参数密钥来防止潜在的恶意攻击。

Rack Utils对65536的密钥空间设置了一个限制，如果您使用大型参数集发送HTTP请求，该密钥空间可能不够大。

您可以通过添加以下到Rails的初始化文件更改此限制：

if Rack::Utils.respond_to?("key_space_limit=") 
    Rack::Utils.key_space_limit = 262144 
end 

我发现这个数字（4倍标准）足够大，以解决H13问题，但对我来说，你可能需要测试不同根据您自己的个人要求进行设置。

这可能不是您特定H13问题的原因，但值得检查作为可能的解决方案。