On Tue, Mar 1, 2016 at 10:33 AM, Rob Brucks <rob.brucks at rackspace.com>
wrote:

> But then why write the slony daemon log file at all?
>
> No timestamps means you have to go look in the syslog file anyway.


I was involved in syslog support, and Steve's thoughts seem
to fit the intent pretty well.

If you use syslog, it independently captures timestamps, and likely more
accurately
than slon logs would (think of the case of aggregating logs across servers;
syslog
would handle coordinating times and merging into one big log).

As Steve mentioned, you wouldn't want to collect two timestamps in each
line,
much preferable to just have one.

The place where we decided to simplify was to have just one representation
to
control contents of log entries, and thus fewer configuration parameters.
In order
to have timestamps in slon's STDOUT/STDERR logs, and leave them out for
syslog, we'd
need more configuration parameters, and it is commonly thought that we've
already got plenty of configuration.

My attitude at the time was that if you decide to request using syslog, then
that presumably was how you wanted to handle monitoring, and that
the STDOUT/STDERR logs become of little importance.

If it's of huge importance to have separate formatting of syslog and native
logging, we could do so.  But my first reaction is to ask why you're keen on
the native logging when syslog lends so much greater flexibility.

We didn't make them mutually exclusive, because it seemed easy
enough to provide both in a form.  But if it takes a lot more sophisticated
configuration to make them useful, well, I'd rather point sophisticated
configuration attentions at doing fancier things with syslog.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://lists.slony.info/pipermail/slony1-bugs/attachments/20160301/03395df6/attachment.htm