Answering my own mail ;)

Ott Köstner wrote:

Still, something locks up in the system, when I increase the number Python threads with http://grub.org/?q=en/node/204
The frequency of these messages increases with 70 threads. And with 200 threads the system locks up totally -- almost no access from web. [I can still access the system and kill the process :) ]

Clearly, there is some limit in the system. It is not memory, cpu or bandwidth. If the number of threads is also not limited, it must be something else. What might it be?

Eventually, I found out that the limiting factor here is named, not capable of handling more request and causing Apache to lock up.

But now I am trapped here. Is there anybody who can help?

In /var/log/messages:
named[63198]: socket: too many open file descriptors

I have googled this, but no help. Recompiled bind, adding -DFD_SETSIZE=2048 to CFLAGS. No help.

sockstat |grep -c named

does not big numbers. With 60 Grubng/Pyhton threads this number is around 20, rises over 100 with 100 threads, but this is still much lower than 2048...

Any ideas?


Testi oma Interneti kiirust / Test Your Internet speed:

freebsd-questions@freebsd.org mailing list
To unsubscribe, send any mail to "freebsd-questions-unsubscr...@freebsd.org"

Reply via email to