> btw do you know if there's anything that would be needed to change if
> syslog-ng were to become a real drop-in replacement for syslogd?
> 

1. Fix syslog-ng config (apart from stupid formatting bug /dev/log seems
to be DGRAM not STREAM contrary to syslog-ng readme).

2. Add proper logrotate to syslog-ng

3. Split sysklogd into klogd and syslogd proper so that you could
replace syslogd part with alternative implementation. You can run
syslog-ng without klogd but then you need /proc/kmsg i.e. it won't work
if /proc is not mounted.

4. Add syslog-ng support to msec.

The reason for a split is, logrotate for syslogd and syslog-ng  refer to
the same files which means logs are rotated twice. What I had in mind is
to make syslog-ng conflict with syslogd. Alternative is (assuming
syslog-ng is adopted) to just check which one is currently running and
prime it. 

Fredl, do you have any plans to add support for external modules? It
would be ideal case - syslog-ng comes with own msec module to avoid
rewriting it every time.

Currently I have fixed RPM (based on syslog-ng 1.5.17); if there is any
interest I can upload it but I do not know what to do with sources - one
file is defined as extra source and AFAIK sources are not entered in CVS
when I upload RPM? Also it does not contain logrotate support for above
reasons.

> And the sysklogd maintainer is .... Warly. you can ask him for the
final
> decision :-)
>

So far syslog-ng seems to work here and it allows _far_ better control
over where your logs go. It has some problems in interaction with klogd
but I bet it is solvable.

-andrej

Reply via email to