[Hawkular-dev] some new hawkular openshift agent stuff

[Gareth Healy]

Hi John,

Just tried deploying latest agent but get an error. Am running the below
steps to deploy the agent:

oc adm policy add-cluster-role-to-user cluster-reader
system:serviceaccount:openshift-infra:hawkular-agent
oc create -f https://raw.githubusercontent.com/hawkular/hawkular-openshift-agent/master/deploy/openshift/hawkular-openshift-agent-configmap.yaml
-n openshift-infra
oc process -f https://raw.githubusercontent.com/hawkular/hawkular-openshift-agent/master/deploy/openshift/hawkular-openshift-agent.yaml
IMAGE_VERSION=latest | oc create -n openshift-infra -f -


But get this error when the agent starts:

E0111 18:34:11.353256       1 node_event_consumer.go:72] Error obtaining
information about the agent pod
[openshift-infra/hawkular-openshift-agent-h1ynb]. err=User
"system:serviceaccount:openshift-infra:hawkular-openshift-agent" cannot get
pods in project "openshift-infra"

I've also got a custom pod that has a configmap mounted, but the agent
never gets past this error message so doesn't collect any metrics.

Output of oc describe on my custom pod shows the volume mount:

Volume Mounts:
      /var/run/configmap/hawkular-agent from hawkular-openshift-agent (rw)

Any ideas?

On Sat, Jan 7, 2017 at 1:44 AM, John Mazzitelli <mazz at redhat.com> wrote:

> FYI: some new things went into HOSA (that's the Hawkular OpenShift Agent
> for the uninitiated).
>
> 1. The agent now emits its own metrics and can monitor itself. Right now
> it just emits some basic "go" metrics like memory usage, CPU usage, etc
> along with one agent-specific one - a counter that counts the number of
> data points it has collected in its lifetime. We'll add more metrics as we
> figure out the things people want to see, but we have the infrastructure in
> place now.
>
> 2. The agent is deployed as a daemonset. This means as new nodes are
> brought online, an agent will go along with it (or so I'm told :)
>
> 3. The agent has changed the way it discovers what to monitor - it no
> longer looks at annotations on pods to determine where the configmaps are
> for those pods. Instead, it looks up volume declarations to see if there is
> an agent configmap defined. This was done to be ready for the future when
> new security constraints will be introduced in OpenShift which would have
> broken our annotation approach. This approach using volumes should not hit
> that issue.
>
> NOTE: If you are building the latest agent from master, we added some
> dependencies so you have to update your dependencies via Glide by using the
> "make update-deps" target prior to building from source.
> _______________________________________________
> hawkular-dev mailing list
> hawkular-dev at lists.jboss.org
> https://lists.jboss.org/mailman/listinfo/hawkular-dev
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://lists.jboss.org/pipermail/hawkular-dev/attachments/20170111/ddd2c085/attachment-0001.html