Nice ! Thanks for the report.
Can we see the inventory in ManageIQ ? ie: does it work end-to-end ?
Can we have a demo ? (Live or recorded)
Thomas
On Fri, Nov 3, 2017 at 12:54 AM, John Mazzitelli <mazz(a)redhat.com> wrote:
The team can report a pretty good milestone as of today. We have
Prometheus linked with Hawkular in a nice way. Not fully integrated (we
don't have, say, a single image with everything) but works pretty good for
now.
Here's some details on how to demo this (you should have hawkular-services
repo git cloned locally and the hawkular-1275 branch checked out):
1) Build hawkular-services dist with -Pdev - we will run it from the
target/ directory in the next step:
mvn clean install -Pdev
2) Run Hawkular-Services:
hawkular-services/dist/target/hawkular-services-dist-*/bin/standalone.sh
-b 0.0.0.0
3) Run Prometheus configured to scan our hawkular metrics endpoints - we
have a test script that runs a pre-configured Prometheus in a docker
container:
hawkular-services/dist/src/test/scripts/run-prometheus.sh
OK, that's it. You are ready to see inventory and metrics.
At this point your server is running and an agent is running with it. We
have an agent attached to h-services so the agent can monitor the server.
The agent itself has a prometheus metrics endpoint exposed via built-in
jmx exporter. You can see the raw data from the agent if you point your
browser to
http://localhost:9779/metrics.
After you give it some time, the agent will report all inventory up to the
server. The server will also create a Prometheus scrape configuration file
to tell Prometheus where the agent's metrics endpoint is. The
pre-configured Prometheus discovers any new scrape configuration files as
new agents come online. So without doing anything manually, Prometheus will
immediately start collecting metrics data from agents as they come online.
You can see the inventory in our little inventory UI - point your browser
to
http://localhost:8080/hawkular/inventory/ui and look at the tree of
resources.
Now for the Prometheus integration - click on a resource in the inventory
UI's left-hand tree, pick a resource that has metrics (say, "[WildFly
Server] Local"). On the right side, you'll see the default base URL for
prometheus listed as "http://localhost:9090" in a text field - you can
leave that as-is or edit it if you need to. But this is referring to the
Prometheus you started in step 3 above. Now drop down the Metrics menu and
pick a metric. The browser opens a tab and takes you to the metric graph in
Prometheus. You should see data.
What does all of this get you? Well, normally Prometheus data isn't
organized by associated resources - its just a long list of labeled metrics
that you have to make sense out of in order to organize the metrics
yourself. Hawkular gives you a nice inventory view (your hierarchy of
resources) that does this organization of Prometheus metrics for you. So
while Prometheus gives you metric data and pretty graphs, Hawkular
"organizes" those metrics by grouping them with their associated resources
in hierarchical inventory view.
--John Mazz
_______________________________________________
hawkular-dev mailing list
hawkular-dev(a)lists.jboss.org
https://lists.jboss.org/mailman/listinfo/hawkular-dev