如何调试hhvm服务器

Question

I have wordpress running in a hiphop vm on ubuntu 12.04. My server works fine for about an hour and then it stops serving requests and just times out. I can see that the hhvm process is still running when I execute ps -aux | grep "hhvm" ps -aux | grep "hhvm" but it is just not responding to any requests. Is there anyway that I can debug this? No errors are appearing in my log file so I have no Idea where to go from here. Is there any way to debug your php code while running in hhvm?

UPDATE:

I uped the threads from 50 to 300 and the server lasted 24 hours and now the same thing is occurring again.

output from the admin page /check-health is

{
  "load":240
,  "queued":0
,  "hhbc-roarena-capac":0
,  "tc-size":17473376
,  "tc-stubsize":12588591
,  "targetcache":37280
,  "units":223
}

and /stats.html is

{
  "stats": {
    "pages": [
    ]
  }
}

Answer 1

It sounds like there might be a bug in the version of HHVM that you are using where the threads gradually hang over time until there are no more worker threads available to serve incoming requests.

You can collect some details by attaching to the process with gdb (by typing "gdb -p $PID" where $PID is hhvm's process ID) and then typing "thread apply all bt" to get the stack traces of all the threads. If you open a github issue at https://github.com/facebook/hiphop-php/issues and provide the stack traces of all the threads someone from the HHVM team should respond and help dig into the issue. The more details you provide on how to reproduce the issue (ex. giving a small, complete code example that reproduces the problem), the more likely it will be that the bug can be identified and fixed.

Also, if you using are an older version of HHVM you might try upgrading to a newer version, since there have been a number of bug fixes over the past few months. I don't know off hand if there were any bug fixes for threads getting stuck, but it's possible upgrading might fix your issue.