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(a)lists.jboss.org https://lists.jboss.org/mailman/listinfo/hawkular-dev

Gareth, I've seen this before and when I did it is because the agent doesn't have the cluster-reader role. So something happened with the command: oc adm policy add-cluster-role-to-user cluster-reader system:serviceaccount:openshift-infra:hawkular-agent I can tell you just recently I changed names to ensure all names in HOSA are consistent (before, you would see some things named "hawkular-agent" and other times you'd see "hawkular-openshift-agent" and other times you'd see things named "openshift-agent" - I made all names consistent... "hawkular-openshift-agent"). One of the changes is the name of the service account - it is now "hawkular-openshift-agent" not "hawkular-agent" So if you are running newer code, check that. Your command above should be: oc adm policy add-cluster-role-to-user cluster-reader system:serviceaccount:openshift-infra:hawkular-openshift-agent Without that role, the agent will not work at all. You can see the Makefile we use to deploy in our dev environments - see the "openshift-deploy" target here: https://github.com/hawkular/hawkular-openshift-agent/blob/ master/Makefile#L42-L46 ----- Original Message ----- > 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(a)redhat.com&gt; 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(a)lists.jboss.org > > https://lists.jboss.org/mailman/listinfo/hawkular-dev > > >