[Labs-l] Simple way to monitor status of tools for tool authors

[Antoine Musso]

Le 22/07/13 21:45, Yuvi Panda a écrit :
> As I put up more tools on toollabs, I find myself wanting monitoring
> more and more to alert me when something related to my tool is down.
> While there can be current solutions hacked up by various methods
> (mine mentions me on github *shudder*), I think having a simple
> solution that's puppetized and usable in a simple way by everyone is
> something we should have.
> 
> I've written up a simple page listing requirements for such a service,
> and will perhaps add implementation details once I've read docs more.
> The page is at:
> 
> https://wikitech.wikimedia.org/wiki/User:Yuvipanda/Icinga_for_tools
> 
> I know nothing of monitoring or icinga, so people with more knowledge
> - please chime in! Also if you are a tool author and have more
> requests / features that would be useful that I'm missing, please
> chime in too.

Hello,

There is an icinga instance on labs at:

 http://icinga.wmflabs.org/icinga/

Its configuration is generated by a python script available in
labs/nagios-builder.git

It does a LDAP query to figure out the class applied on an instance and
then associate them with hardcoded monitoring check.

I am not sure that is ideal.  I would prefer us to reuse the
operations/puppet way of adding monitor with monitor_service /
monitor_host etc.  This way people reusing the existing manifests will
benefit monitoring out of the box.


-- 
Antoine "hashar" Musso