Bug #1652
closedLighttpd Occasionally Crashes on Cygwin Windows Server 2003 R2
Description
I got CYGWIN_NT-5.2 on Windows Server 2003 R2.
Hardware: 4 x Intel(R) Xeon(TM) CPU 3.00GHz (probably 2 dual cores) with 4GB of RAM.
$ lighttpd -V
lighttpd-1.4.19 - a light and fast webserver
Build-Date: Mar 30 2008 18:42:34
Event Handlers:
+ select (generic)
+ poll (Unix)
- rt-signals (Linux 2.4+)
- epoll (Linux 2.6)
- /dev/poll (Solaris)
- kqueue (FreeBSD)Network handler:
+ writev
+ mmap supportFeatures:
- IPv6 support
+ zlib support
+ bzip2 support
+ crypt support
- SSL Support
+ PCRE support
- mySQL support
- LDAP support
- memcached support
- FAM support
- LUA support
- xml support
- SQLite support
- GDBM supportOccasionally it crashes minutes after staring, with:
2008-04-25 11:49:38: (connections.c.1210) connection closed: poll() -> ERR 6Alex
Files
Updated by stbuehler over 16 years ago
That message just means that a connection got closed, as the system reported an error for it.
I don't think lighty crashes just as a connection got closed, so can you perhaps provide more info, e.g. a backtrace?
Updated by alevchuk over 16 years ago
Thanks!
I haven't been getting the crashes lately, but I did not change anything to fix that, so the problem is probably still there. I do not think that running lighty as a child of strace for many days would be a good idea. But if I find out what causes the crash then I will give you the system call trace.
The problem may be related to the lack of memory, because I had one other service. It was more verbose and left a "cannot malloc" in my logs.
With the method explained by Cygwin I found that no more then 1G can be malloc'ed on my system. More info is here:
http://www.cygwin.com/cygwin-ug-net/setup-maxmem.html
My initial default was 1G, and not 384 MB as said in the article. I was not able to increase my heap limit using the instructions that they provided.
Attached is my config.
Updated by alevchuk over 16 years ago
Here are the references of people with similar problems:
Updated by stbuehler over 16 years ago
You don't need to run it in strace, gdb / valgrind is enough for a backtrace.
Updated by alevchuk over 16 years ago
I installed gdb 6.8.0 from the package and valgrind has will not "./confugure" in cygwin (asks me to consider building a port).
The server crashed again today (once).
Can you tell me which lines of lighttpd-1.4.19 to break point so I can run under in gdb for several days?
Updated by stbuehler over 16 years ago
gdb should break if lighttpd "fails" (segfaults), so you shouldn't need break points and just running it in gdb should be enough.
Ah, and to get as much information as possible, please try the full backtrace:
bt full
Updated by admin over 16 years ago
That message just means that a connection got closed, as the system reported an error for it.
Are you sure? Lighttpd doesn't log all connections being closed, so I'm sure it's more than 'just' a connection being closed.
Updated by stbuehler over 16 years ago
- Status changed from New to Fixed
- Resolution set to worksforme
Iirc, it is a close after a connection reset.
Missing feedback.
Updated by stbuehler about 16 years ago
- Status changed from Fixed to Missing Feedback
Also available in: Atom