On Fri, 2005-10-14 at 14:08 -0400, Elliot Lee wrote:
On Thu, 13 Oct 2005, Stephen J. Smoogen wrote:
> One of the things we have to do at our sites is remove the Red
> Hat/Fedora syslog services and replace them with syslog-ng. The TCP
> part, the better message congestion issues, and the better ability to
> group logging together is really needed by larger sites. [We have
> found that the lossiness of regular syslog can get black marks on
> reviews.] A long time ago, it was proposed for inclusion around Red
> Hat Linux 6.0 but was considered too unstable and some other issues.
>
> Could it be evaluated again for either for replacing the older syslog
> system or working the syslog system so that alternatives could cover
> it with syslog-ng in extras? What issues would need to be done to help
> this along?
There are enough syslog replacement wannabees out there that this isn't a
slam-dunk decision. Because the client-side interface is built into glibc,
so many apps use the syslog interface that just putting syslog-ng into the
distro won't accomplish much for admins.
(For syslog-ng in particular, I seem to recall when I looked at it that it
avoided making radical-but-necessary improvements for the sake of
compatibility, which makes it sound a bit like subversion to me.)
I think someone needs to evaluate the millions of syslog replacements out
there, pick the best one overall, and get the client-side API for that
into glibc. Then it'll just be a matter of starting to convert apps from
the old API to the new one.
okay I'm confused. Why isn't using syslog-ng a slam-dunk.
1. we don't break compatibility for the api - which means not modifying
lots of difficult programs
2. it makes it easier for this to trickle into rhel
3. it's a known value in that it's used in large-scale and small-scale
operations all over the place.
4. it's an immediate win over syslog in that it is:
a. more feature-full
b. actually maintained
-sv