StackLight provides reduced list of metrics.

[Dmitry Kabanov]

MOS 8.0. StackLight Version 0.10-0.10.0-1.

Steps to reproduce:
1. Deploy a cloud without LMA (3 controllers, 2 computes).
2. Install LMA plugins. Configure LMA plugins.
3. Add 1(3) node(s) and assign LMA roles.
4. Deploy changes.
5. Check Grafana, Kibana and Nagios.

Actual result:
Metrics for OpenStack services was not provided in Grafana and Kibana. Almost all services were in UNKNOWN state in Nagios.

Expected result:
All metrics for all OpensStack services are present.

[Simon Pasquier (simon-pasquier)]

Please attach the Fuel diagnostic snapshot + the output of the LMA diagnostic command [1].

[1] http://fuel-plugin-lma-collector.readthedocs.io/en/latest/configure_plugin.html#diagnostic-tool

[Dmitry Kabanov (dkabanov)]

Added Fuel diagnostic snapshot (fuel-snapshot-2016-08-09_23-45-44.tar.xz).

[Dmitry Kabanov (dkabanov)]

But I have a problem with LMA diagnostic tool. The LMA diagnostic seems to be stuck (I executed the command from fuel master node):

"[root@nailgun ~]# bash /var/www/nailgun/plugins/lma_collector-0.10/contrib/tools/diagnostic.sh
Running lma_diagnostic tool on all nodes which are ready and online (this can take several minutes)

"

I decided to run diagnostic on all nodes. And I see the same output on all non-LMA nodes:

"root@node-1:~# lma_diagnostics
lma_diagnostics: command not found"

You can find the output of the lma_diagnostics command from LMA node(s) in lma_diagnostic.log file.

[Swann Croiset (swann-w)]

for other nodes (controllers, computes, .) , the command should be there: /usr/local/bin/lma_diagnostics
can you check it exists and run it please.

[Swann Croiset (swann-w)]

the lma_diagnostics store diagnostic information in this directory:

/var/lma_diagnostics/

[Dmitry Kabanov (dkabanov)]

Thank you for your response. Unfortunately, I was not able to run the command. I can't find it:

root@node-1:~# find /usr/. -name 'lma_diagnostics'
root@node-1:~#

[Simon Pasquier (simon-pasquier)]

There's a limitation for post-deployment with MOS 8 that I forgot to mention although it's described in the documentation [1]:

> The StackLight Collector Plugin is a hot-pluggable plugin. Therefore, it is possible to install and
> deploy the collector in an environment that is already deployed. After the installation of the
> StackLight Collector Plugin, define the settings of the plugin and run the command shown below from
> the Fuel master node for every node of your deployment starting with the controller node(s):
>
> [root@nailgun ~]# fuel nodes --env <env_id> --node <node_id> --start \
> post_deployment_start --tasks hiera

Please run the command above and tell us if it fixes your issue.

[1] http://fuel-plugin-lma-collector.readthedocs.io/en/latest/configure_plugin.html#plugin-configuration

[Dmitry Kabanov (dkabanov)]

The command was run on all nodes except nodes with LMA. Unfortunately, the command finished unsuccessfully. And the cluster is in error state right now.

You can find the error message below:
"Could not find data item lma::collector::node_profiles in any Hiera data file and no default supplied at /etc/fuel/plugins/lma_collector-0.10/puppet/manifests/configure_apt.pp:33 on node node-1.test.domain.local"

[Simon Pasquier (simon-pasquier)]

The documentation has been updated to address this issue: https://review.openstack.org/#/c/353560/