[racket-dev] Suggestions for monitoring unresponsive web server connection?

From: Neil Van Dyke (neil at neilvandyke.org)
Date: Tue Sep 6 20:35:21 EDT 2011

Kathi Fisler wrote at 09/06/2011 08:13 PM:
> Are there commands we can use when we startup racket or the server
> that might give diagnostics to help trace the problem?
>   

Intermittent failures are a headache.  In addition to whatever people 
advise here, you might want to add your own detailed logging to a file, 
with timestamps.  Don't be afraid to dump dozens of lines of log output 
per HTTP request.

> We are not getting core dumps (even when the process dies).

You also might want to confirm that your kernel is set up to do core 
dumps, that the process limits on the actual "racket" process would 
permit core dumps, and that whatever directory&file the kernel config 
would say to put the core dump would be writable.

All these are about arranging to capture diagnostic info, should an 
intermittent failure occur again.  In addition to that, you might want 
to audit the code, looking for potential deadlocks.  If you're using the 
FFI or C extensions, you have to assess how much you trust the code and 
whether you're calling it in a safe way, since auditing C code will be 
harder than auditing Racket code.

Oh, and Matthew discovered a conceivable issue with address space 
randomization, and I don't know whether he's 100% ruled it out as a 
problem.  You might want to disable that feature in Linux, just in case 
(although there are reasons why they added that feature, but I think the 
feature is less relevant if the only server code you're running on the 
machine is in Racket).

-- 
http://www.neilvandyke.org/



Posted on the dev mailing list.