[chef] Re: Re: Re: Re: Re: Best practice for measuring and monitoring chef-client runs?

[Augie Schwer < >]

Thanks everyone, that is all very helpful.

On Wed, Sep 17, 2014 at 12:04 PM, Tiago Cruz < " target="_blank"> > wrote:

I have a simple check on nagios like this

...

HOURS=2

SECONDS=$(expr $HOURS \* 60 \* 60)

OHAI_TIME="$(expr $(date +%s) - $SECONDS)"

SEARCH="$(knife search node "ohai_time:[* TO $OHAI_TIME] AND chef_environment:production")"

...

On Wed, Sep 17, 2014 at 3:35 PM, DV < " target="_blank"> > wrote:

We have a custom Rails app that acts as handler for chef-client. Here's what dashboard looks like: http://i.imgur.com/sR4UCWC.png

We also have an automated task that runs "knife status" and reports on any hosts that haven't checked in for a while.

On Wed, Sep 17, 2014 at 9:37 AM, Mark Mzyk < " target="_blank"> > wrote:

The report handler that supplies the data from the client run to the Chef server reporting add on is open source, so it could be used and/or built off of, if you didn't want to use the pre-built Chef add ons.

It's here in the client: https://github.com/opscode/chef/blob/master/lib/chef/resource_reporter.rb

- Mark Mzyk

" style="color:#737f92!important;padding-right:6px;font-weight:bold;text-decoration:none!important" target="_blank">Steffen Gebert

September 17, 2014 at 2:03 AM

Jumping into the "we do.." postings: We send chef-client statistics to
zabbix using a report handler:
- success
- elapsed_time
- start_time
- end_time
- all_resources_num
- updated_resource num

I think that should be pretty easy to adapt to whatever monitoring
system you use.

Yours
Steffen

Links:
https://github.com/TYPO3-cookbooks/zabbix-custom-checks/blob/master/recipes/chef-client.rb
https://github.com/TYPO3-cookbooks/zabbix-custom-checks/blob/master/templates/default/chef-client/chef-client-handler.rb

" style="color:#737f92!important;padding-right:6px;font-weight:bold;text-decoration:none!important" target="_blank">Augie Schwer

September 16, 2014 at 8:33 PM

What are people using to monitor and measure their chef-client runs?

I would like to monitor for when chef-client runs fail on a node.

It would be nice to measure chef-client run times.

Is it safe to assume people are using handlers for both of these? What are some popular ways to accomplish these goals? Thanks!

--
Augie Schwer    -    " target="_blank">     -    http://schwer.us

--
Best regards, Dmitriy V.

--
-- Tiago Cruz

--
Augie Schwer    -      -    http://schwer.us