On Wed, Mar 11, 2009 at 08:20:03AM +0100, Balazs Scheidler wrote:
So from now on, it is possible to do things like:
source aaa { udp(facility(syslog) level(emerg)); };
And if the message does not have a proper syslog header containing the above values, it'll use the ones specified in the configuration file.
Now as I think of it, it might be better to really use a separate keyword for this, e.g. default-facility() and default-level() Any opinions?
Hi, Thanks ! I think it's better to have a separate keyword for this case.
Hmm.. I was thinking about this feature a bit more, and it occurred to me that it would be possible to implement this also as a rewrite function, e.g.
source src { ... };
rewrite r_fac { set("syslog" value("FACILITY")); set("crit" value("LEVEL")); };
I'm not sure that we need both features: * to specify the default priority value for any kind of log source * to make it possible to rewrite the value
I find the 2nd more general. What do you all think?
I had not though about the rewrite thing, but it sounds good too. The two features look good to me. At first I just wanted to set a value facility/level to messages who did not have one. But it could be usefull to have the more general one, with the rewrite rule. I just hope it is not a performance penalty to call a rewrite rule instead of just setting the facility/level when the LogMessage structure is created. By the way, talking about performance, I have started looking for using polling instead of follow-freq(). I'll really have many source log files to check. The more files are to be checked, the more stat() and lseek() are called by syslog-ng. It could be a performance gain to use poll() or epoll() (on linux at least for this one) on the source files to check for new messages. It seems that polling is only applicable on /proc/kmsg for now. Am I wrong or did I miss something (or both !) ? Regards. -- Xavier