Christoffer Hall-Frederiksen wrote:
> > > Two things. It should not be sorted in to files but sent as local?.* to
> > > a loghost using syslog.
> >
> > Then have the filter program use 'logger' to re-log the messages.
> >
> > > And what happens if the program listening on the
> > > other end of the fifo dies?
> >
> > Ensure that it gets re-started.
>
> This is a very nasty solution, a potential problem could very well be
> that the kernel writes to the fifo and fills the buffer before the
> program kan read it (or (re)start and read it). Wouldn't that make
> the write block? Generally I'm not too fond of the kernel beeing
> dependant on a perticular program running. There must be another
> way of doing it!
1. If you specify a FIFO in syslog.conf, then it's syslogd that is
going to be writing to it, not the kernel.
2. I presume that syslogd will simply discard the data if the FIFO is
full (the alternative would be for syslogd to block, which would
result in a lot more data being discarded).
3. Logging kernel messages to file already depends upon klogd reading
data out of the kernel's ring buffer before it overruns.
--
Glynn Clements <[EMAIL PROTECTED]>
-
To unsubscribe from this list: send the line "unsubscribe linux-net" in
the body of a message to [EMAIL PROTECTED]
