[ale] Hardware monitoring

[Chris Ricker]

On Mon, 22 May 2006, Christopher Fowler wrote:

> Nagios is mostly GUI.  You still have to do much work to get it to
> monitor everything.  This is what I saw when I installed it 6 months
> ago.  Many people have been releasing their monitoring scripts.   
> 
> The problem with monitoring over 1000 servers is that you really do not
> want anything that does mostly polling.  Yea you can poll for things
> like services, icmp, etc.  But you do not want to poll for things like
> disk usage, temp, anything internal.  In those cases you need to install
> an agent that can report back via SNMP when those problems exists.  I'm
> not aware of any OSS products that have an SNMP agent piece that can be
> installed on a Linux, Sun, Windows, etc server.  

You'd be surprised -- I know there are several sites using hobbit to do 
monitoring of thousands of hosts. But then, hobbit started out as a Big 
Brother rewrite just to improve performance for large sites, before it 
became its own thing....

munin can be configured in either push or poll or combo push / pull 
models, so it might also be an option

snmppd + nagios might be suitable as well, if going all / mostly snmp is 
possible

Some of the other open source stuff may scale that high as well, but those 
are the ones I'd start considering depending upon what features are 
needed....

later,
chris