[syslog-ng] UDP packet loss with syslog-ng

[Clayton Dukes]

Hi Lars,
I'm sorry if my response didn't help you, however, I have been through
exactly what you are describing and it is outlined in that link I sent.
Please check the "update" part of the wiki that quotes what Baszi told me
and why I ended up using the recvbuf in syslog-ng. Setting it there made
everything work for me.




______________________________________________________________

Clayton Dukes
______________________________________________________________


On Fri, Oct 15, 2010 at 5:11 PM, Lars Kellogg-Stedman <lars at oddbit.com>wrote:

> > Actually, you did not say it in your message. Because searching it for
> > the string 'buf' returns no results.
>
> Look, I'm not trying to be difficult, but I do at least expect people
> to read my email before responding.  Here, I'll quote it for you:
>
>  If I raise the rmem settings like this:
>
>   net.core.rmem_default = 512000
>   net.core.rmem_max = 1024000
>
>  Then it looks like I can support messages rates around 1000 msgs/sec.
>  If I try with 2000 msgs/sec, the loss rates jumps up again (to around
>  30%).
>
> I'm sorry if you got bored before reading it.  I was trying to provide
> sufficient details that people wouldn't suggest the things I've
> already tried.
>
> Additionally, as far as I can tell, setting net.core.rmem_default is
> largely equivalent, from the point of view of syslog-ng, to setting
> so_rcvbuf() (since it sets the default receive buffer size).  On a
> system with lots of open sockets, setting a lower
> net.core.rmem_default and using so_rcvbuf() would probably make more
> sense.
>
> ______________________________________________________________________________
> Member info: https://lists.balabit.hu/mailman/listinfo/syslog-ng
> Documentation:
> http://www.balabit.com/support/documentation/?product=syslog-ng
> FAQ: http://www.campin.net/syslog-ng/faq.html
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://lists.balabit.hu/pipermail/syslog-ng/attachments/20101015/0588af73/attachment-0001.htm