cvs commit: src/etc/defaults periodic.conf

[Robert Watson]

On Tue, 31 Jan 2006, Anton Yuzhaninov wrote:

> RW> My daily script parsers certainly aren't.  I quite like being able to pull in
> RW> a mailbox of old daily output and plot disk space use over time.
>
> For automated monitoring better (and more safe in case of changing daily 
> mail output) to use SNMP based tools like nagios, zabbix, remstats, cacti...
>
> Daily mail more useful for human reading.

Actually, the daily (etc) script output proves to be a very reliable and 
useful way to manage long-term monitoring.  The format is generally consistent 
(xcept when commits like the one that spawned this thread happen), and the 
logs are easily kept locally or reliably sent over the network.  If the 
network is down, sendmail happily spools the output for later delivery, it can 
be sent to multiple recipients without special tools, and so on.  I recognize 
the usefulness of SNMP, but think SNMP is a tool you use to support things 
like long-term monitoring, not the presentation piece itself.  One of the 
nicest things about the periodic scripts is that they require literally zero 
configuration -- you don't have to set up daemons, manage keying material, 
maintain additional packages, etc.  With recent bsnmp work, SNMP has become a 
lot more capable of monitoring the interesting things about FreeBSD, but it's 
still not what I'm actually looking for.

I don't know if anyone has written it yet, but it would be quite neat to have 
an SNMP-driven implementation of {top, netstat, ps, df, ifconfig} that can be 
easily pointed at bsnmpd on local or remote machines.  That way you get the 
current UNIX management interface, which is well-understood and easily 
scripted, with a distributed back-end.

Robert N M Watson