Re: Monitoring

Le Mardi 20 D=E9cembre 2005 09:59, Buchan Milne a =E9crit=A0:
> I wonder if it would be useful discusssing monitoring, and which monitori=
ng
> tools we should concentrate more on, and possibly try and get one of them
> into main (it seems weird that we have no monitoring tool in main ...).
>
> As some of you know, I currently work for an ISP. We have two separate
> deployments, with about 35 production servers running RHEL in one, and
> about 16 production servers running RHEL in the other. We deploy the
> servers using kickstart files generated from a configuration database, so
> that all packages, configurations etc for the host are done during
> installation (which in the case of machines with lights-out management can
> be done/initiated from anywhere that has network access).
>
> (A number of packages we run on RHEL are rebuilds of Mandriva packages,
> including a number I maintain, such as OpenLDAP, hobbit etc).
>
> We are currently running Nagios as the monitoring tool for the larger
> deployment, but due to a number of reasons, we have decided to run Hobbit
> (a BigBrother clone) as the monitoring tool for the smaller deployment, a=
nd
> have also set Hobbit up for the larger deployment. This (non-production)
> Hobbit installation for the larger deployment is accessible at present, at
> http://196.25.211.20/hobbit/ .
>
> Anyway, some of the reasons we are using Hobbit are:
>
> -does both status monitoring/alerting and trend monitoring
> -integrated trend monitoring of all native checks (disk, cpu, memory, tcp
> service response times)
> -built-in ssl certificate checks for any ssl-enabled service (ie https)
> -a large collection of additional checks (from http://www.deadcat.net),
> including all BigBrother extension scripts. For example, we monitor the HP
> Insight Manager snmp data via the CIM extension).
> -ease of writing extensions, and being able to monitor trends in the
> results of those extensions (via the ncv rrd plugin), for example:
> http://196.25.211.20/hobbit-cgi/bb-hostsvc.sh?HOSTSVC=3Dio.ol which uses =
this
> script I wrote: http://www.zarb.org/~bgmilne/bb-openldap.pl
> -less complexity in the interface than Nagios, while supporting (AFAICS)
> all the functionality
> -less complexity in configuration (ie configuration for service monitoring
> for each hosts consists of one line in the bb-hosts file)
>
>
> Other tools that I haven't really had a chance to look at in much detail:
>
> 1)monitorix, which Antoine uploaded recently
> http://www.monitorix.org
> Seems to have most of the features of Hobbit.
>
> 2)Oreon
> http://www.oreon-project.org/
>
> Seems to be a better Nagios frontend, addresses some of the problems I ha=
ve
> with Nagios.
>
> 3)Zabbix
> http://www.zabbix.org/
>
> Also has win32 clients.
>
> 4)Cacti
> AFAIK, just trend monitoring using rrd+MySQL
>
> So, should we package all the ones we don't have yet, or decide on featur=
es
> that are necessary, and package/support only one or two?

IMHO we should package all the ones we don't have yet. However having a wiz=
ard=20
or an integrated solution allowing to setup the server and its monitoring=20
could be usefull. Actually you have to do it manually.

during a moment I was thinking about IMC=20
( http://imc.sourceforge.net/home.html ) as this is a mix between webmin +=
=20
nagios + cacti, but the dev seems to be stalled and it's a very complicated=
=20
architecture.

=2D-=20
(=E9crit dans le livre d'or de plusieurs restaurants parisiens)
Je m'ai bien r=E9galer. signe: Marguerite Duras
Pierre Desproges.