On 23 January 2010 11:10, Jake Anderson <[email protected]> wrote:
> I believe redirect stdout to /dev/null but leave stderror
> if something errors you get the error message, but normal output just goes
> into the bit bucket

That depends a lot on the programs actually behaving correctly, also
you might want to log correct execution so you know if/when something
stopped working, and what was executed when the error occured.

We drowned in e-mail to one mailbox until we executed a project to:

1. Replace e-mails by logging into log files.
2. Define logwatch to know when things are wrong in each log file.
3. configure logrotate
4. Setup a separate mailbox system to receive all the e-mails from the
system, file them into a shared IMAP folder tree by source (e.g.
script X on system Y).

We are still developing/tweaking this system but so far it already
helped in cleaning up our inboxes to some degree. Eventually I hope to
receive only e-mails containing genuine errors and maybe some weekly
summaries.

Another challenge is to concentrate logwatch from dozens of hosts into
one report, instead of wading through dozens of e-mails. I'm still
trying to figure out a good solution to that.  Our attempt to use
rsyslog to concentrate all logs from all systems on a single server
caused the application to slow down much more than we could live with,
so for now we are stuck with direct write's to local files.

--Amos
-- 
SLUG - Sydney Linux User's Group Mailing List - http://slug.org.au/
Subscription info and FAQs: http://slug.org.au/faq/mailinglists.html

Reply via email to