Hello!
I've just created ASP.Net website and ran it on Apache&mono on Linux.
And I've got strange situation. After 50.000-100.000 requests mod-mono-server
process unexpectedly become busy with 100% or 200% (exactly) CPU load.
After restarting mono process via mono-ctl panel cpu consumption returns to
1-3%.
I've found some errors like
[Wed Sep 03 16:02:57 2014] [error] (32)Broken pipe: write_data failed
[Wed Sep 03 16:02:57 2014] [error] (32)Broken pipe: write_data failed
[Wed Sep 03 16:02:57 2014] [error] command failed: failed to get client block
(data)
in apache error_log, but it seems like there are no relations in time between
these errors and busy moment.
So, the questions:
1) are there any way to know what is happening in this situation? What threads
are running, etc... I'm not sure the reason is in mono itself, may be the reason
is in my application. Full profiling on production server is not possible, I
mean just a few statistics,
2) if the reason is in garbage collector: are there any ways to tune it? Or to
switch from sgen to old Boehm or so on?
Any help are highly appreciated!
P.S. Mono 3.2.8 compiled from sources on standard CentOS 6.5
# mono -V
Mono JIT compiler version 3.2.8 (tarball Wed Jun 11 00:48:42 MSK 2014)
Copyright (C) 2002-2014 Novell, Inc, Xamarin Inc and Contributors.
www.mono-project.com
TLS: __thread
SIGSEGV: altstack
Notifications: epoll
Architecture: amd64
Disabled: none
Misc: softdebug
LLVM: supported, not enabled.
GC: sgen
# httpd -v
Server version: Apache/2.2.15 (Unix)
Server built: Apr 3 2014 23:56:16
# uname -a
Linux web1.server 2.6.32-431.17.1.el6.x86_64 #1 SMP Wed May 7 23:32:49 UTC 2014
x86_64 x86_64 x86_64 GNU/Linux
With best wishes,
Max Karavaev
_______________________________________________
Mono-list maillist - [email protected]
http://lists.ximian.com/mailman/listinfo/mono-list
