summaryrefslogtreecommitdiffstats
path: root/docs/platform
diff options
context:
space:
mode:
Diffstat (limited to 'docs/platform')
-rw-r--r--docs/platform/feature_statemgmt.rst32
-rw-r--r--docs/platform/index.rst2
-rw-r--r--docs/platform/srmPdpxPdpMgmt.pngbin76009 -> 54563 bytes
-rw-r--r--docs/platform/srmPdpxResiliencyPdpMgmt1.pngbin58164 -> 55150 bytes
-rw-r--r--docs/platform/srmPdpxResiliencyPdpMgmt2.pngbin57400 -> 55096 bytes
-rw-r--r--docs/platform/srmPdpxScalingPdpMgmt1.pngbin0 -> 58893 bytes
-rw-r--r--docs/platform/swarch_srm.rst624
7 files changed, 556 insertions, 102 deletions
diff --git a/docs/platform/feature_statemgmt.rst b/docs/platform/feature_statemgmt.rst
index 70b0f186c..960ff396d 100644
--- a/docs/platform/feature_statemgmt.rst
+++ b/docs/platform/feature_statemgmt.rst
@@ -7,22 +7,22 @@ Feature: State Management
*************************
.. contents::
- :depth: 3
+ :depth: 2
Summary
^^^^^^^
The State Management Feature provides:
- - Node-level health monitoring
- - Monitoring the health of dependency nodes - nodes on which a particular node is dependent
- - Ability to lock/unlock a node and suspend or resume all application processing
- - Ability to suspend application processing on a node that is disabled or in a standby state
- - Interworking/Coordination of state values
- - Support for ITU X.731 states and state transitions for:
- - Administrative State
- - Operational State
- - Availability Status
- - Standby Status
+- Node-level health monitoring
+- Monitoring the health of dependency nodes - nodes on which a particular node is dependent
+- Ability to lock/unlock a node and suspend or resume all application processing
+- Ability to suspend application processing on a node that is disabled or in a standby state
+- Interworking/Coordination of state values
+- Support for ITU X.731 states and state transitions for:
+ - Administrative State
+ - Operational State
+ - Availability Status
+ - Standby Status
Usage
^^^^^
@@ -153,6 +153,16 @@ Node Health Monitoring
If the forward progress counter stops incrementing, the integrity monitoring routine will assume the node application has lost sanity and it will trigger a *statechange* (disableFailed) to cause the operational state to become disabled and the availability status attribute to become failed. Once the forward progress counter again begins incrementing, the operational state will return to enabled.
+**Application Monitoring with AllSeemsWell**
+
+ The IntegrityMonitor class provides a facility for applications to directly control updates of the forwardprogressentity table. As previously described, *startTransaction()* and *endTransaction()* are provided to monitor the forward progress of transactions. This, however, does not monitor things such as internal threads that may be blocked or die. An example is the feature-state-management *DroolsPdpElectionHandler.run()* method.
+
+ The *run()* method is monitored by a timer task, *checkWaitTimer()*. If the *run()* method is stalled an extended period of time, the *checkWaitTimer()* method will call *StateManagementFeature.allSeemsWell(<className>, <AllSeemsWell State>, <String message>)* with the AllSeemsWell state of Boolean.FALSE.
+
+ The IntegrityMonitor instance owned by StateManagementFeature will then store an entry in the allSeemsWellMap and block updates of the forwardprogressentity table. This in turn, will cause the Drools PDP operational state to be set to “disabled” and availability status to be set to “failed”.
+
+ Once the blocking condition is cleared, the *checkWaiTimer()* will again call the *allSeemsWell()* method and include an AllSeemsWell state of Boolean.True. This will cause the IntegrityMonitor to remove the entry for that className from the allSeemsWellMap and allow updating of the forwardprogressentity table, so long as there are no other entries in the map.
+
**Dependency Monitoring**
When a Drools PDP (or other node using the *IntegrityMonitor* policy/common module) is dependent upon other nodes to perform its function, those other nodes can be defined as dependencies in the properties file. In order for the dependency algorithm to function, the other nodes must also be running the *IntegrityMonitor*. Periodically the Drools PDP will check the state of dependencies. If all of a node type have failed, the Drools PDP will declare that it can no longer function and change the operational state to disabled and the availability status to dependency.
diff --git a/docs/platform/index.rst b/docs/platform/index.rst
index 4baf57e90..3dca9ec16 100644
--- a/docs/platform/index.rst
+++ b/docs/platform/index.rst
@@ -20,6 +20,7 @@ Policy Installation and Deployment
installationVM.rst
installAmsterController.rst
installBeijingController.rst
+ swarch_srm.rst
deployment.rst
Policy Software Architecture
@@ -28,7 +29,6 @@ Policy Software Architecture
.. toctree::
:maxdepth: 1
- swarch_srm.rst
swarch_pdp.rst
feature_eelf.rst
feature_testtransaction.rst
diff --git a/docs/platform/srmPdpxPdpMgmt.png b/docs/platform/srmPdpxPdpMgmt.png
index 5a998f080..69360f470 100644
--- a/docs/platform/srmPdpxPdpMgmt.png
+++ b/docs/platform/srmPdpxPdpMgmt.png
Binary files differ
diff --git a/docs/platform/srmPdpxResiliencyPdpMgmt1.png b/docs/platform/srmPdpxResiliencyPdpMgmt1.png
index 84d468fe3..50b3ee86e 100644
--- a/docs/platform/srmPdpxResiliencyPdpMgmt1.png
+++ b/docs/platform/srmPdpxResiliencyPdpMgmt1.png
Binary files differ
diff --git a/docs/platform/srmPdpxResiliencyPdpMgmt2.png b/docs/platform/srmPdpxResiliencyPdpMgmt2.png
index 6b5ae8510..46721f3ae 100644
--- a/docs/platform/srmPdpxResiliencyPdpMgmt2.png
+++ b/docs/platform/srmPdpxResiliencyPdpMgmt2.png
Binary files differ
diff --git a/docs/platform/srmPdpxScalingPdpMgmt1.png b/docs/platform/srmPdpxScalingPdpMgmt1.png
new file mode 100644
index 000000000..f04cbee4f
--- /dev/null
+++ b/docs/platform/srmPdpxScalingPdpMgmt1.png
Binary files differ
diff --git a/docs/platform/swarch_srm.rst b/docs/platform/swarch_srm.rst
index 2de6baebc..1483eb24e 100644
--- a/docs/platform/swarch_srm.rst
+++ b/docs/platform/swarch_srm.rst
@@ -2,14 +2,14 @@
.. This work is licensed under a Creative Commons Attribution 4.0 International License.
.. http://creativecommons.org/licenses/by/4.0
-*****************************************
-Scalability, Resiliency and Manageability
-*****************************************
+*************
+Policy on OOM
+*************
.. contents::
:depth: 3
-The new Beijing release scalability, resiliency, and manageablity are described here. These capabilities apply to the OOM/Kubernetes installation.
+The new Beijing release capabilities for OOM are described here.
Installation
^^^^^^^^^^^^
@@ -18,7 +18,9 @@ Follow the OOM installation instructions at http://onap.readthedocs.io/en/latest
Overview of the running system
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
-Upon initialization, you should see pools of 4 PDP-Ds and 2 PDP-Xs:
+Upon initialization, you should see the following pods, one instance for each Policy component: PAP, PDP-X, BRMSGW, PDP-D, policydb, and nexus.
+
+Note the "-0" suffix for PDP-X and PDP-D components, which will be increased as they are scaled out to improve runtime performance and reliability.
.. code-block:: bash
:caption: verify pods
@@ -27,17 +29,13 @@ Upon initialization, you should see pools of 4 PDP-Ds and 2 PDP-Xs:
onap dev-brmsgw-5dbc4c8dc4-llk5s 1/1 Running 0 18m 10.42.120.43 k8sx
onap dev-drools-0 1/1 Running 0 18m 10.42.60.27 k8sx
- onap dev-drools-1 1/1 Running 0 16m 10.42.105.190 k8sx
- onap dev-drools-2 1/1 Running 0 15m 10.42.139.82 k8sx
- onap dev-drools-3 1/1 Running 0 15m 10.42.128.4 k8sx
onap dev-nexus-7d96568f5f-qp5td 1/1 Running 0 18m 10.42.172.8 k8sx
onap dev-pap-8587696769-vwj6k 2/2 Running 0 18m 10.42.19.137 k8sx
onap dev-pdp-0 2/2 Running 0 18m 10.42.144.218 k8sx
- onap dev-pdp-1 2/2 Running 0 15m 10.42.233.111 k8sx
onap dev-policydb-587d55bdff-4f5dz 1/1 Running 0 18m 10.42.12.242 k8sx
-and a service for every component:
+You will also see a service for every component:
.. code-block:: bash
:caption: verify services
@@ -51,28 +49,37 @@ and a service for every component:
onap pdp ClusterIP None <none> 8081/TCP 24m
onap policydb ClusterIP None <none> 3306/TCP 24m
-Config and Decision policy requests will be distributed across PDP-Xs through the *pdp* service. PDP-X clients (such as DCAE) should configure their URLs to go through the *pdp* service. Their requests will be distributed across the available PDP-X replicas. The PDP-Xs can be also accessed individually (dev-pdp-0 and dev-pdp-1 above), but is preferable to that external clients use the service.
+Config and Decision policy requests will be distributed across PDP-Xs through the *pdp* service. PDP-X clients (such as DCAE) should configure their URLs to go through the *pdp* service. Their requests will be distributed across the available PDP-X replicas.
+The PDP-Xs can be also accessed individually (dev-pdp-0, or dev-pdp-x if scaled out), but is preferable for PDP-X external clients to interface through the service.
-PDP-Ds are also accessible on a group fashion by using the service IP. Nevertheless, as DMaaP is the main means of communication with other ONAP components, the service interface is not used heavily.
+PDP-Ds are also accessible on a group fashion by using the service IP, but DMaaP is the main means of communication with other ONAP components.
Healthchecks
^^^^^^^^^^^^
-Verify that the policy healtcheck passes by the robot framework:
+Verify that the policy healthcheck passes by the robot framework:
.. code-block:: bash
- :caption: ~/oom/kubernetes/robot/ete-k8s.sh onap health 2> /dev/null | grep PASS
+ :caption: robot healthcheck
+ ~/oom/kubernetes/robot/ete-k8s.sh onap health 2> /dev/null | grep PASS
+ ..
Basic Policy Health Check | PASS |
+ ..
-A policy healthcheck (with more detailed output) can be done directly to the drools service in the policy VM.
+
+A policy healthcheck (with more detailed output) can be done directly by invoking the drools service in the policy VM.
.. code-block:: none
- :caption: Healtcheck on the PDP-D service
+ :caption: PDP-D service (more detailed) healthcheck
- curl --silent --user '<username>:<password> -X GET http://localhost:30217/healthcheck | python -m json.tool
+ # Using default credentials for the healtcheck service.
+ # To change the default username and passwords for this service,
+ # please modify configuration pre-installation at:
+ # oom/kubernetes/policy/charts/drools/resources/config/opt/policy/config/drools/keys/feature-healthcheck.conf
+ curl --silent --user 'healthcheck:zb!XztG34' -X GET http://localhost:30217/healthcheck | python -m json.tool
{
"details": [
{
@@ -99,17 +106,17 @@ A policy healthcheck (with more detailed output) can be done directly to the dro
],
"healthy": true
}
+
-
-PDP-X active/active pool
+PDP-X Active/Active Pool
^^^^^^^^^^^^^^^^^^^^^^^^
-The policy engine UI (console container in the pap pod) can be used to check that the the 2 individual PDP-Xs are synchronized.
-The console URL is accessible at ``http://<oom-vm>:30219/onap/login.htm``. Select the PDP tab.
+The policy engine UI (console container in the pap pod) can be used to check that the PAP and the PDP-Xs are synchronized.
+The console URL is accessible at ``http://<oom-vm>:30219/onap/login.htm``. Select the *PDP* menu entry on the left side panel under *Policy*.
.. image:: srmPdpxPdpMgmt.png
-After initialization, there will be no policies loaded into the policy subsystem. You can verify it, by accessing the Editor tab in the UI.
+After initialization, there will be no policies loaded into the policy subsystem. This can be verified by accessing the Editor tab in the UI.
PDP-D Active/Active Pool
@@ -122,8 +129,13 @@ The following command can be issued on each of the PDP-D replicas IPs:
.. code-block:: bash
:caption: Querying the rules association for a PDP-D replica
- curl --silent --user '<username>:<password>' -X GET http://<drools-replica-ip>:9696/policy/pdp/engine/controllers/amsterdam/drools | python -m json.tool
-
+ # Using default credentials for the drools telemetry service.
+ # To change the default username and passwords for this service,
+ # please modify configuration pre-installation at:
+ # oom/kubernetes/policy/charts/drools/resources/config/opt/policy/config/drools/base.conf
+
+
+ curl --silent --user '@1b3rt:31nst31n' -X GET http://<drools-replica-ip>:9696/policy/pdp/engine/controllers/amsterdam/drools | python -m json.tool
{
"alive": false,
"artifactId": "NO-ARTIFACT-ID",
@@ -138,18 +150,56 @@ The following command can be issued on each of the PDP-D replicas IPs:
"version": "NO-VERSION"
}
+
+
+Before Installing Policies
+^^^^^^^^^^^^^^^^^^^^^^^^^^
+
+It has been experienced in large OOM k8s multi-node full ONAP installations that components DNS and connectivity problems across pods through services. Eventually, the system becomes stable and ready to be used. Single node, smaller installations, do not seem to have these issues. Give the system enough time to make sure it has been initialized properly before pushing policies.
+
+Make sure the policy subsystem is initialized by:
+
+ 1. Verify that the "PDP Management" screen shows the 1 pooled PDP-X "UP_TO_DATE". If the PDP-X does not show the correct state, restart the faulty one to force re-synchronization with the pap.
+
+ .. code-block:: bash
+ :caption: Force re-synchronization of a PDP-X
+
+ kubectl exec -it dev-pdp-0 --container pdp -n onap -- bash -c "source /opt/app/policy/etc/profile.d/env.sh; policy.sh stop; policy.sh start"
+
+ # bounce the BRMSGW as well since it synchronizes with PDP-Xs via websockets:
+
+ kubectl exec -it dev-brmsgw-b877bc567-wbnbz -n onap -- bash -c "source /opt/app/policy/etc/profile.d/env.sh; policy.sh stop; policy.sh start"
+
+
+ 2. Verify service name resolution is OK across policy components
+
+ .. code-block:: bash
+ :caption: Verify policy services connectivity
+
+ # pick any policy pod to run these tests from:
+ # kubectl get pods --all-namespaces -o=wide
+
+ kubectl exec -it dev-brmsgw-b877bc567-wbnbz -n onap -- bash -c "ping policydb"
+ kubectl exec -it dev-brmsgw-b877bc567-wbnbz -n onap -- bash -c "ping pdp"
+ kubectl exec -it dev-brmsgw-b877bc567-wbnbz -n onap -- bash -c "ping drools"
+ kubectl exec -it dev-brmsgw-b877bc567-wbnbz -n onap -- bash -c "ping nexus"
+ kubectl exec -it dev-brmsgw-b877bc567-wbnbz -n onap -- bash -c "ping message-router"
+
+
Installing Policies
^^^^^^^^^^^^^^^^^^^
-The OOM default installation will come with no policies pre-configured. There is a sample script used by integration teams to load policies to support all 4 use cases at: /tmp/policy-install/config/push-policies.sh in the pap container within the pap pod. This script can be modified for your own particular installation, for example if only interested in vCPE use cases, remove those vCPE related API REST calls. For the vFW use case, you may want to edit the encoded operational policy to point to the proper resourceID in your installation.
+The OOM default installation will come with no policies pre-configured. There is a sample script used by integration teams to load policies to support all four use cases at: */tmp/policy-install/config/push-policies.sh* in the pap container within the pap pod. This script can be modified for your own particular installation, for example if only interested in vCPE use cases, remove those vCPE related API REST calls. For the vFW use case, you may want to edit the encoded operational policy to point to the proper resourceID in your installation.
-The above mentioned push-policies.sh script can be executed as follows:
+The above mentioned *push-policies.sh* script can be executed as follows:
.. code-block:: bash
:caption: Installing the default policies
- kubectl exec -it dev-pap-8587696769-vwj6k -c pap -n onap -- bash -c "export PRELOAD_POLICIES=true; /tmp/policy-install/config/push-policies.sh"
+ # NOTE: If modifications are required to the /tmp/policy-install/config/push-policies.sh, it should be copied
+ # to a different location, for example /tmp as /tmp/policy-install/config directory is read-only.
+ kubectl exec -it dev-pap-8587696769-vwj6k -c pap -n onap -- bash -c "export PRELOAD_POLICIES=true; /tmp/policy-install/config/push-policies.sh"
..
Create BRMSParam Operational Policies
@@ -189,22 +239,29 @@ The above mentioned push-policies.sh script can be executed as follows:
..
Transaction ID: 6d47db63-7956-4f5f-ab34-aeb5a124a90d --Policy 'com.Config_MS_MicroServicevCPE.1.xml' was successfully pushed to the PDP group 'default'.
-
-The policies pushed can be viewed through the Policy UI:
+The policies pushed could be viewed eventually through the Policy UI:
.. image:: srmEditor.png
-As a consequence of pushing the policies, the brmsgw component will compose drools rules artifacts and publish them to the nexus respository at ``http://<oom-vm>:30236/nexus/``
+As part of the process pushing of policies through the policy, the brmsgw component will compose drools rules artifacts and publish them to the nexus respository at ``http://<oom-vm>:30236/nexus/``.
.. image:: srmNexus.png
-At the same time each replica of the PDP-Ds will receive notifications for each new version of the policies to run for the amsterdam controller. You can run the following command to see how the amsterdam controller is associated with the latest rules version. The following command can be used for verification for each replica:
+At the same time each replica of the PDP-Ds will receive notifications for each new version of the policies to run for the Amsterdam controller. The following command can be run to see how the amsterdam controller is associated with the latest rules version.
+
+The following command can be used for verifying each replica:
.. code-block:: none
:caption: Querying the rules association of a PDP-D replica
- curl --silent --user '<username><password> -X GET http://<replica-ip>:9696/policy/pdp/engine/controllers/amsterdam/drools | python -m json.tool
+ # Using default credentials for the drools telemetry service.
+ # To change the default username and passwords for this service,
+ # please modify configuration pre-installation at:
+ # oom/kubernetes/policy/charts/drools/resources/config/opt/policy/config/drools/base.conf
+
+ curl --silent --user '@1b3rt:31nst31n' -X GET http://<replica-ip>:9696/policy/pdp/engine/controllers/amsterdam/drools | python -m json.tool
+
{
"alive": true,
"artifactId": "policy-amsterdam-rules",
@@ -223,13 +280,20 @@ At the same time each replica of the PDP-Ds will receive notifications for each
"version": "0.4.0"
}
-Likewise, for verification purposes, each PDP-X replica can be queried directly to retrieve policy information. The following commands can be used to query a policy through the pdp service:
+Likewise, for verification purposes, each PDP-X replica can be queried directly to retrieve policy information.
+
+The following commands can be used to query a policy through the pdp service:
.. code-block:: bash
:caption: Querying the "pdp" service for the vFirewal policy
+ # Open a shell into the pap pod
+
ubuntu@k8sx:~$ kubectl exec -it dev-pap-8587696769-vwj6k -c pap -n onap bash
+
+ # In this example the vFirewall policy is queried.
+
policy@dev-pap-8587696769-vwj6k:/tmp/policy-install$ curl --silent -X POST --header 'Content-Type: application/json' --header 'Accept: application/json' --header 'ClientAuth: cHl0aG9uOnRlc3Q=' --header 'Authorization: Basic dGVzdHBkcDphbHBoYTEyMw==' --header 'Environment: TEST' -d '{"policyName": ".*vFirewall.*"}' http://pdp:8081/pdp/api/getConfig | python -m json.tool
[
{
@@ -270,12 +334,18 @@ Likewise, for verification purposes, each PDP-X replica can be queried directly
]
-while the following commands could be used to query an specific PDP-X replica:
+While the following commands could be used to query an specific PDP-X replica:
.. code-block:: bash
:caption: Querying PDP-X 0 for the vCPE policy
+ # open a shell into the pap pod
+
+ ubuntu@k8sx:~$ kubectl exec -it dev-pap-8587696769-vwj6k -c pap -n onap bash
+
+ # in this example the vCPE policy is queried.
+
curl --silent -X POST --header 'Content-Type: application/json' --header 'Accept: application/json' --header 'ClientAuth: cHl0aG9uOnRlc3Q=' --header 'Authorization: Basic dGVzdHBkcDphbHBoYTEyMw==' --header 'Environment: TEST' -d '{"policyName": ".*vCPE.*"}' http://10.42.144.218:8081/pdp/api/getConfig | python -m json.tool
[
{
@@ -318,19 +388,32 @@ while the following commands could be used to query an specific PDP-X replica:
PDP-X Resiliency
^^^^^^^^^^^^^^^^
-A PDP-X container failure can be simulated by performing a"policy.sh stop" operation within the PDP-X container, this in fact will shutdown the PDP-X service. The kubernetes liveness operation will detect that the ports are down, inferring there's a problem with the service, and in turn, will restart the container. In the following example will cause PDP-X 1 to fail.
+A PDP-X container failure can be simulated by either:
+ a) performing a"policy.sh stop" operation within the PDP-X container, which in fact will shutdown the PDP-X service, and eventually will be detected by the liveness checks, or
+ b) by plainly deleting the corresponding pod.
+
+In the following example, the PDP-X 0 is forced to fail.
.. code-block:: bash
- :caption: Causing PDP-X 1 service to fail
+ :caption: Causing PDP-X 0 service to fail
- ubuntu@k8sx:~$ kubectl exec -it dev-pdp-1 --container pdp -n onap -- bash -c "source /opt/app/policy/etc/profile.d/env.sh; policy.sh stop;"
+ # In these scenarios the liveness check will fail and recovery actions will take place.
+
+
+ # Alternative 1: In this scenario we shutdown the PDP-X 0 service, so the liveness monitored ports will be down
+ # (but the pod is up) and corrective measures will be applied
+ ubuntu@k8sx:~$ kubectl exec -it dev-pdp-0 --container pdp -n onap -- bash -c "source /opt/app/policy/etc/profile.d/env.sh; policy.sh stop;"
pdplp: STOPPING ..
pdp: STOPPING ..
+
+ # Alternative 2: Brute force delete of the PDP-X 0 pod.
+ ubuntu@k8sx:~$ kubectl delete pod dev-pdp-0 -n onap
+ pod "dev-pdp-0" deleted
-Upon detection of the service being down through the liveness check, the container will be restarted. Note the restart count when querying the status of the pods:
+Upon detection of the service being down through the liveness check, the container will be restarted. Note the **restart count** when querying the status of the pods:
.. code-block:: bash
- :caption: Checking PDP-X 1 restart count
+ :caption: Checking PDP-X 0 restart count
ubuntu@k8sx:~$ kubectl get pods --all-namespaces -o=wide
@@ -338,30 +421,29 @@ Upon detection of the service being down through the liveness check, the contain
onap dev-brmsgw-5dbc4c8dc4-llk5s 1/1 Running 0 3d 10.42.120.43 k8sx
onap dev-drools-0 1/1 Running 0 3d 10.42.60.27 k8sx
- onap dev-drools-1 1/1 Running 0 3d 10.42.105.190 k8sx
- onap dev-drools-2 1/1 Running 0 3d 10.42.139.82 k8sx
- onap dev-drools-3 1/1 Running 0 3d 10.42.128.4 k8sx
onap dev-nexus-7d96568f5f-qp5td 1/1 Running 0 3d 10.42.172.8 k8sx
onap dev-pap-8587696769-vwj6k 2/2 Running 0 3d 10.42.19.137 k8sx
onap dev-pdp-0 2/2 Running 0 3d 10.42.144.218 k8sx
- onap dev-pdp-1 2/2 Running 1 3d 10.42.233.111 k8sx <--- **
onap dev-policydb-587d55bdff-4f5dz 1/1 Running 0 3d 10.42.12.242 k8sx
-During the restart process, the PAP component, will detect that PDP-X 1 is down and therefore its state being reflected in the PDP-X screen:
+During the restart process, the PAP component, will detect that PDP-X 0 is down and therefore its state being reflected in the PDP-X screen:
.. image:: srmPdpxResiliencyPdpMgmt1.png
-This screen will be updated to reflect PDP-X 1 is back alive, after PDP-X 1 synchronizes itself with the PAP.
+This screen will be updated to reflect PDP-X 0 is back alive, after PDP-X 0 synchronizes itself with the PAP.
.. image:: srmPdpxResiliencyPdpMgmt2.png
-At that point, PDP-X is usable either directly or through the service to query for policies.
+At that point, the PDP-X is usable either directly or through the service to query for policies.
.. code-block:: bash
:caption: Query PDP-X 1 for vCPE policy
+ # in this example we perform the vCPE query from the OOM VM
+ # the default installation credentials are used for querying the vCPE policy
+
ubuntu@k8sx:~$ curl --silent -X POST --header 'Content-Type: application/json' --header 'Accept: application/json' --header 'ClientAuth: cHl0aG9uOnRlc3Q=' --header 'Authorization: Basic dGVzdHBkcDphbHBoYTEyMw==' --header 'Environment: TEST' -d '{"policyName": ".*vCPE.*"}' http://10.42.233.111:8081/pdp/api/getConfig | python -m json.tool
[
{
@@ -404,12 +486,21 @@ At that point, PDP-X is usable either directly or through the service to query f
PDP-D Resiliency
^^^^^^^^^^^^^^^^
-A PDP-D container failure can be simulated by performing a"policy stop" operation within the PDP-D container, this in fact will shutdown the PDP-D service. The kubernetes liveness operation will detect that the ports are down, inferring there's a problem with the service, and in turn, will restart the container. In the following example will cause PDP-D 3 to fail.
+A PDP-D container failure can be simulated by either:
+ a) performing a"policy stop" operation within the PDP-D pod, which in fact will shutdown the PDP-D service, and eventually will be detected by the liveness checks, or
+ b) by plainly deleting the corresponding pod.
+
+In the following example, the PDP-D 0 is forced to fail.
.. code-block:: bash
- :caption: Causing PDP-D 3 to fail
+ :caption: Causing PDP-D 0 to fail
+
+ # In these scenarios the liveness check will fail and recovery actions will take place.
- ubuntu@k8sx:~/oom/kubernetes$ kubectl exec -it dev-drools-3 --container drools -n onap -- bash -c "source /opt/app/policy/etc/profile.d/env.sh; policy stop"
+ # Alternative 1: in this scenario we shutdown the PDP-D 0 policy process, so the liveness monitored ports
+ # will be down (but the pod is up) and corrective measures will be applied
+
+ ubuntu@k8sx:~/oom/kubernetes$ kubectl exec -it dev-drools-0 --container drools -n onap -- bash -c "source /opt/app/policy/etc/profile.d/env.sh; policy stop"
[drools-pdp-controllers]
L []: Stopping Policy Management... Policy Management (pid=3284) is stopping... Policy Management has stopped.
@@ -417,79 +508,432 @@ A PDP-D container failure can be simulated by performing a"policy stop" operatio
Upon detection of the service being down through the liveness check, the container will be restarted. Note the restart count when querying the status of the pods:
.. code-block:: bash
- :caption: Checking PDP-D 3 restart count
+ :caption: Checking PDP-D 0 restart count
- ubuntu@k8sx:~/oom/kubernetes$ kubectl get pods --all-namespaces -o=wide
-
- NAMESPACE NAME READY STATUS RESTARTS AGE IP NODE
+ ubuntu@k8sx:~$ kubectl get pods --all-namespaces -o=wide | grep drool
+ onap dev-drools-0 0/1 Running 0 1d 10.42.10.21 k8sx
+ ..
+ ubuntu@k8sx:~$ kubectl get pods --all-namespaces -o=wide | grep drools
+ onap dev-drools-0 1/1 Running 1 1d 10.42.10.21 k8sx <-- note restart count
+
+Verification that the restarted PDP-D 0 comes up with the appropriate policy loaded can be verified by checking its maven coordinates:
+
+.. code-block:: bash
+ :caption: Verifying restarted PDP-D points to policies pre-failure.
+
+ ubuntu@k8sx:~$ curl --silent --user '@1b3rt:31nst31n' -X GET http://10.42.10.21:9696/policy/pdp/engine/controllers/amsterdam/drools | python -m json.tool
+ {
+ "alive": true,
+ "artifactId": "policy-amsterdam-rules",
+ "brained": true,
+ "groupId": "org.onap.policy-engine.drools.amsterdam",
+ "locked": false,
+ "modelClassLoaderHash": 189820624,
+ "recentSinkEvents": [],
+ "recentSourceEvents": [],
+ "sessionCoordinates": [
+ "org.onap.policy-engine.drools.amsterdam:policy-amsterdam-rules:0.4.0:closedloop-amsterdam"
+ ],
+ "sessions": [
+ "closedloop-amsterdam"
+ ],
+ "version": "0.4.0"
+ }
- onap dev-brmsgw-5549d99466-7989k 1/1 Running 0 1h 10.42.252.245 k8sx
- onap dev-drools-0 1/1 Running 0 1h 10.42.30.52 k8sx
- onap dev-drools-1 1/1 Running 0 1h 10.42.9.245 k8sx
- onap dev-drools-2 1/1 Running 0 1h 10.42.95.0 k8sx
- onap dev-drools-3 1/1 Running 1 1h 10.42.224.52 k8sx
- onap dev-nexus-6558979c95-xlxcc 1/1 Running 0 1h 10.42.142.36 k8sx
- onap dev-pap-64b67f66b9-lc8vl 2/2 Running 0 1h 10.42.187.255 k8sx
- onap dev-pdp-0 2/2 Running 0 1h 10.42.164.57 k8sx
- onap dev-pdp-1 2/2 Running 0 1h 10.42.155.145 k8sx
- onap dev-policydb-7d4b75869-qd8n5 1/1 Running 0 1h 10.42.148.37 k8sx
-
PDP-X Scaling
^^^^^^^^^^^^^
-To scale a new PDP-X, set the replica count appropriately. In our scenario below, we are going to scale the PDP-X with a new replica, PDP-X 2, to have a pool of 3 PDP-X.
+To scale a new PDP-X, set the replica count appropriately.
+
+In our tests below, we are going to work with the OOM policy component in isolation. In this exercise, we scale the PDP-X with 1 additional replica, PDP-X 1.
.. code-block:: bash
:caption: Scaling a PDP-X
- helm upgrade -i dev local/onap --namespace onap --set policy.pdp.replicaCount=3
-
+ ubuntu@k8sx:~$ helm upgrade -i dev local/onap --namespace onap --set global.pullPolicy=IfNotPresent --set policy.pdp.replicaCount=2
Release "dev" has been upgraded. Happy Helming!
- LAST DEPLOYED: Mon May 14 01:37:03 2018
+ LAST DEPLOYED: Mon Jun 4 15:19:05 2018
NAMESPACE: onap
STATUS: DEPLOYED
- ..
- kubectl get pods --all-namespaces -o=wide
+ RESOURCES:
+ ==> v1/Service
+ NAME TYPE CLUSTER-IP EXTERNAL-IP PORT(S) AGE
+ dbc-pg-primary ClusterIP 10.43.29.226 <none> 5432/TCP 2d
+ dbc-pg-replica ClusterIP 10.43.202.168 <none> 5432/TCP 2d
+ dbc-postgres ClusterIP 10.43.181.134 <none> 5432/TCP 2d
+ dmaap-bc NodePort 10.43.254.230 <none> 8080:30241/TCP,8443:30242/TCP 2d
+ message-router-kafka ClusterIP 10.43.69.159 <none> 9092/TCP 2d
+ message-router-zookeeper ClusterIP None <none> 2181/TCP 2d
+ message-router NodePort 10.43.123.102 <none> 3904:30227/TCP,3905:30226/TCP 2d
+ msb-consul NodePort 10.43.27.77 <none> 8500:30285/TCP 2d
+ msb-discovery NodePort 10.43.178.20 <none> 10081:30281/TCP 2d
+ msb-eag NodePort 10.43.77.235 <none> 80:30282/TCP,443:30284/TCP 2d
+ msb-iag NodePort 10.43.221.196 <none> 80:30280/TCP,443:30283/TCP 2d
+ brmsgw NodePort 10.43.21.222 <none> 9989:30216/TCP 2d
+ nexus NodePort 10.43.159.27 <none> 8081:30236/TCP 2d
+ drools NodePort 10.43.233.67 <none> 6969:30217/TCP,9696:30221/TCP 2d
+ policydb ClusterIP None <none> 3306/TCP 2d
+ pdp ClusterIP None <none> 8081/TCP 2d
+ pap NodePort 10.43.110.50 <none> 8443:30219/TCP,9091:30218/TCP 2d
+ robot NodePort 10.43.172.248 <none> 88:30209/TCP 2d
- NAMESPACE NAME READY STATUS RESTARTS AGE IP NODE
- ..
- onap dev-pdp-0 2/2 Running 0 1h 10.42.164.57 k8sx
- onap dev-pdp-1 2/2 Running 0 1h 10.42.155.145 k8sx
- onap dev-pdp-2 2/2 Running 0 1m 10.42.47.58 k8sx
- ..
+ ==> v1beta1/Deployment
+ NAME DESIRED CURRENT UP-TO-DATE AVAILABLE AGE
+ dev-dmaap-bus-controller 1 1 1 1 2d
+ dev-message-router-kafka 1 1 1 1 2d
+ dev-message-router-zookeeper 1 1 1 1 2d
+ dev-message-router 1 1 1 1 2d
+ dev-kube2msb 1 1 1 1 2d
+ dev-msb-consul 1 1 1 1 2d
+ dev-msb-discovery 1 1 1 1 2d
+ dev-msb-eag 1 1 1 1 2d
+ dev-msb-iag 1 1 1 1 2d
+ dev-brmsgw 1 1 1 1 2d
+ dev-nexus 1 1 1 1 2d
+ dev-policydb 1 1 1 1 2d
+ dev-pap 1 1 1 1 2d
+ dev-robot 1 1 1 1 2d
+
+ ==> v1beta1/StatefulSet
+ NAME DESIRED CURRENT AGE
+ dev-dbc-pg 2 2 2d
+ dev-drools 1 1 2d
+ dev-pdp 2 2 2d
+
+ ==> v1/PersistentVolumeClaim
+ NAME STATUS VOLUME CAPACITY ACCESS MODES STORAGECLASS AGE
+ dev-message-router-kafka Bound dev-message-router-kafka 2Gi RWX 2d
+ dev-message-router-zookeeper Bound dev-message-router-zookeeper 2Gi RWX 2d
+ dev-nexus Bound dev-nexus 2Gi RWX 2d
+ dev-policydb Bound dev-policydb 2Gi RWX 2d
+
+ ==> v1/ConfigMap
+ NAME DATA AGE
+ dev-dmaap-bus-controller-config 1 2d
+ dev-message-router-cadi-prop-configmap 1 2d
+ dev-message-router-msgrtrapi-prop-configmap 1 2d
+ dev-msb-discovery 1 2d
+ dev-msb-eag 1 2d
+ dev-msb-iag 1 2d
+ dev-brmsgw-pe-configmap 2 2d
+ dev-drools-configmap 6 2d
+ dev-drools-log-configmap 1 2d
+ dev-drools-settings-configmap 1 2d
+ dev-policydb-configmap 1 2d
+ dev-pdp-log-configmap 1 2d
+ dev-pdp-pe-configmap 3 2d
+ dev-pe-scripts-configmap 1 2d
+ dev-filebeat-configmap 1 2d
+ dev-pe-configmap 1 2d
+ dev-pap-pe-configmap 7 2d
+ dev-pap-sdk-log-configmap 1 2d
+ dev-pap-log-configmap 1 2d
+ dev-robot-resources-configmap 3 2d
+ dev-robot-lighttpd-authorization-configmap 1 2d
+ dev-robot-eteshare-configmap 4 2d
+
+ ==> v1/PersistentVolume
+ NAME CAPACITY ACCESS MODES RECLAIM POLICY STATUS CLAIM STORAGECLASS REASON AGE
+ dev-dbc-pg-data0 1Gi RWO Retain Bound onap/dev-dbc-pg-data-dev-dbc-pg-0 dev-dbc-pg-data 2d
+ dev-dbc-pg-data1 1Gi RWO Retain Bound onap/dev-dbc-pg-data-dev-dbc-pg-1 dev-dbc-pg-data 2d
+ dev-message-router-kafka 2Gi RWX Retain Bound onap/dev-message-router-kafka 2d
+ dev-message-router-zookeeper 2Gi RWX Retain Bound onap/dev-message-router-zookeeper 2d
+ dev-nexus 2Gi RWX Retain Bound onap/dev-nexus 2d
+ dev-policydb 2Gi RWX Retain Bound onap/dev-policydb 2d
+
+ ==> v1beta1/ClusterRoleBinding
+ NAME AGE
+ onap-binding 2d
+
+ ==> v1/Pod(related)
+ NAME READY STATUS RESTARTS AGE
+ dev-dmaap-bus-controller-5bd859c7dc-blzdc 1/1 Running 0 2d
+ dev-message-router-kafka-748cdf7b9c-srv7l 1/1 Running 0 2d
+ dev-message-router-zookeeper-5b5969f6f-8rk9w 1/1 Running 0 2d
+ dev-message-router-b5bdc599c-5h56k 1/1 Running 0 2d
+ dev-kube2msb-579fc77c54-m84qx 1/1 Running 0 2d
+ dev-msb-consul-7bc4fcc8-94gsc 1/1 Running 0 2d
+ dev-msb-discovery-768547bcb-2hr7j 2/2 Running 0 2d
+ dev-msb-eag-5d95686c67-9lkzs 2/2 Running 0 2d
+ dev-msb-iag-675b649848-pv2gh 2/2 Running 0 2d
+ dev-brmsgw-5675f5877b-wv68s 1/1 Running 0 2d
+ dev-nexus-7d96568f5f-m8c4l 1/1 Running 0 2d
+ dev-policydb-587d55bdff-9gdjv 1/1 Running 0 2d
+ dev-pap-678b44cd87-wxbww 2/2 Running 0 2d
+ dev-robot-589c76bb6b-hrrdn 1/1 Running 0 2d
+ dev-dbc-pg-0 1/1 Running 0 2d
+ dev-dbc-pg-1 1/1 Running 0 2d
+ dev-drools-0 1/1 Running 1 2d
+ dev-pdp-0 2/2 Running 1 2d
+ dev-pdp-1 0/2 Init:0/1 0 0s
+
+ ==> v1/Secret
+ NAME TYPE DATA AGE
+ dev-dbc-pg Opaque 3 2d
+ dev-message-router-secret Opaque 1 2d
+ dev-drools-secret Opaque 2 2d
+ dev-policydb-secret Opaque 2 2d
+ onap-docker-registry-key kubernetes.io/dockercfg 1 2d
+
+
+Check Policy Engine UI how the PDP-Xs are coming up and request policies to the PAP.
+
+Eventually the new PDP-X will be connected and serving policies:
+
+ .. image:: srmPdpxScalingPdpMgmt1.png
+
+The new PDP-X should be now ready to serve policies:
+
+.. code-block:: bash
+ :caption: Check that the new PDP-X 3 and 4 can serve policies
+
+ ubuntu@k8sx:~/oom/kubernetes$ curl --silent -X POST --header 'Content-Type: application/json' --header 'Accept: application/json' --header 'ClientAuth: cHl0aG9uOnRlc3Q=' --header 'Authorization: Basic dGVzdHBkcDphbHBoYTEyMw==' --header 'Environment: TEST' -d '{"policyName": ".*vCPE.*"}' http://10.42.183.0:8081/pdp/api/getConfig | python -m json.tool
+ [
+ {
+ "config": ..
+ "matchingConditions": {
+ "ConfigName": "BRMS_PARAM_RULE",
+ "ONAPName": "DROOLS"
+ },
+ "policyConfigMessage": "Config Retrieved! ",
+ "policyConfigStatus": "CONFIG_RETRIEVED",
+ "policyName": "com.Config_BRMS_Param_BRMSParamvCPE.1.xml",
+ "policyType": "BRMS_PARAM",
+ "policyVersion": "1",
+ "property": null,
+ "responseAttributes": {
+ "controller": "amsterdam"
+ },
+ "type": "OTHER"
+ },
+ {
+ "config": "{\"service\":\"tca_policy\",\"location\":\"SampleServiceLocation\",\"uuid\":\"test\",\"policyName\":\"MicroServicevCPE\",\"description\":\"MicroService vCPE Policy\",\"configName\":\"SampleConfigName\",\"templateVersion\":\"OpenSource.version.1\",\"version\":\"1.1.0\",\"priority\":\"1\",\"policyScope\":\"resource=SampleResource,service=SampleService,type=SampleType,closedLoopControlName=ControlLoop-vCPE-48f0c2c3-a172-4192-9ae3-052274181b6e\",\"riskType\":\"SampleRiskType\",\"riskLevel\":\"1\",\"guard\":\"False\",\"content\":{\"tca_policy\":{\"domain\":\"measurementsForVfScaling\",\"metricsPerEventName\":[{\"eventName\":\"Measurement_vGMUX\",\"controlLoopSchemaType\":\"VNF\",\"policyScope\":\"DCAE\",\"policyName\":\"DCAE.Config_tca-hi-lo\",\"policyVersion\":\"v0.0.1\",\"thresholds\":[{\"closedLoopControlName\":\"ControlLoop-vCPE-48f0c2c3-a172-4192-9ae3-052274181b6e\",\"version\":\"1.0.2\",\"fieldPath\":\"$.event.measurementsForVfScalingFields.additionalMeasurements[*].arrayOfFields[0].value\",\"thresholdValue\":0,\"direction\":\"EQUAL\",\"severity\":\"MAJOR\",\"closedLoopEventStatus\":\"ABATED\"},{\"closedLoopControlName\":\"ControlLoop-vCPE-48f0c2c3-a172-4192-9ae3-052274181b6e\",\"version\":\"1.0.2\",\"fieldPath\":\"$.event.measurementsForVfScalingFields.additionalMeasurements[*].arrayOfFields[0].value\",\"thresholdValue\":0,\"direction\":\"GREATER\",\"severity\":\"CRITICAL\",\"closedLoopEventStatus\":\"ONSET\"}]}]}}}",
+ "matchingConditions": {
+ "ConfigName": "SampleConfigName",
+ "Location": "SampleServiceLocation",
+ "ONAPName": "DCAE",
+ "service": "tca_policy",
+ "uuid": "test"
+ },
+ "policyConfigMessage": "Config Retrieved! ",
+ "policyConfigStatus": "CONFIG_RETRIEVED",
+ "policyName": "com.Config_MS_MicroServicevCPE.1.xml",
+ "policyType": "MicroService",
+ "policyVersion": "1",
+ "property": null,
+ "responseAttributes": {},
+ "type": "JSON"
+ }
+ ]
+
+
+ ubuntu@k8sx:~/oom/kubernetes$ curl --silent -X POST --header 'Content-Type: application/json' --header 'Accept: application/json' --header 'ClientAuth: cHl0aG9uOnRlc3Q=' --header 'Authorization: Basic dGVzdHBkcDphbHBoYTEyMw==' --header 'Environment: TEST' -d '{"policyName": ".*vCPE.*"}' http://10.42.137.241:8081/pdp/api/getConfig | python -m json.tool
+ [
+ {
+ "config": "{\"service\":\"tca_policy\",\"location\":\"SampleServiceLocation\",\"uuid\":\"test\",\"policyName\":\"MicroServicevCPE\",\"description\":\"MicroService vCPE Policy\",\"configName\":\"SampleConfigName\",\"templateVersion\":\"OpenSource.version.1\",\"version\":\"1.1.0\",\"priority\":\"1\",\"policyScope\":\"resource=SampleResource,service=SampleService,type=SampleType,closedLoopControlName=ControlLoop-vCPE-48f0c2c3-a172-4192-9ae3-052274181b6e\",\"riskType\":\"SampleRiskType\",\"riskLevel\":\"1\",\"guard\":\"False\",\"content\":{\"tca_policy\":{\"domain\":\"measurementsForVfScaling\",\"metricsPerEventName\":[{\"eventName\":\"Measurement_vGMUX\",\"controlLoopSchemaType\":\"VNF\",\"policyScope\":\"DCAE\",\"policyName\":\"DCAE.Config_tca-hi-lo\",\"policyVersion\":\"v0.0.1\",\"thresholds\":[{\"closedLoopControlName\":\"ControlLoop-vCPE-48f0c2c3-a172-4192-9ae3-052274181b6e\",\"version\":\"1.0.2\",\"fieldPath\":\"$.event.measurementsForVfScalingFields.additionalMeasurements[*].arrayOfFields[0].value\",\"thresholdValue\":0,\"direction\":\"EQUAL\",\"severity\":\"MAJOR\",\"closedLoopEventStatus\":\"ABATED\"},{\"closedLoopControlName\":\"ControlLoop-vCPE-48f0c2c3-a172-4192-9ae3-052274181b6e\",\"version\":\"1.0.2\",\"fieldPath\":\"$.event.measurementsForVfScalingFields.additionalMeasurements[*].arrayOfFields[0].value\",\"thresholdValue\":0,\"direction\":\"GREATER\",\"severity\":\"CRITICAL\",\"closedLoopEventStatus\":\"ONSET\"}]}]}}}",
+ "matchingConditions": {
+ "ConfigName": "SampleConfigName",
+ "Location": "SampleServiceLocation",
+ "ONAPName": "DCAE",
+ "service": "tca_policy",
+ "uuid": "test"
+ },
+ "policyConfigMessage": "Config Retrieved! ",
+ "policyConfigStatus": "CONFIG_RETRIEVED",
+ "policyName": "com.Config_MS_MicroServicevCPE.1.xml",
+ "policyType": "MicroService",
+ "policyVersion": "1",
+ "property": null,
+ "responseAttributes": {},
+ "type": "JSON"
+ },
+ {
+ "config": ...
+ "ConfigName": "BRMS_PARAM_RULE",
+ "ONAPName": "DROOLS"
+ },
+ "policyConfigMessage": "Config Retrieved! ",
+ "policyConfigStatus": "CONFIG_RETRIEVED",
+ "policyName": "com.Config_BRMS_Param_BRMSParamvCPE.1.xml",
+ "policyType": "BRMS_PARAM",
+ "policyVersion": "1",
+ "property": null,
+ "responseAttributes": {
+ "controller": "amsterdam"
+ },
+ "type": "OTHER"
+ }
+ ]
PDP-D Scaling
^^^^^^^^^^^^^
-To scale a new PDP-D, set the replica count appropriately. In our scenario below, we are going to scale the PDP-D service with a new replica, PDP-D 4, to have a pool of 5 PDP-D.
+To scale a new PDP-D, set the replica count appropriately. In our scenario below, we are going to scale the PDP-D service to add a new pod (2 active PDP-Ds).
.. code-block:: bash
:caption: Scaling a PDP-D
- helm upgrade -i dev local/onap --namespace onap --set policy.drools.replicaCount=5
+ # Note: we also set the PDP-X pool to 2 instances (matching the previous section)
+
+
+ ubuntu@k8sx:~$ helm upgrade -i dev local/onap --namespace onap --set global.pullPolicy=IfNotPresent --set policy.pdp.replicaCount=2 --set policy.drools.replicaCount=2
Release "dev" has been upgraded. Happy Helming!
- LAST DEPLOYED: Mon May 14 01:45:19 2018
+ LAST DEPLOYED: Mon Jun 4 15:52:46 2018
NAMESPACE: onap
STATUS: DEPLOYED
- ubuntu@k8sx:~/oom/kubernetes$ kubectl get pods --all-namespaces -o=wide
- NAMESPACE NAME READY STATUS RESTARTS AGE IP NODE
- ..
- onap dev-drools-0 1/1 Running 0 1h 10.42.30.52 k8sx
- onap dev-drools-1 1/1 Running 0 1h 10.42.9.245 k8sx
- onap dev-drools-2 1/1 Running 0 1h 10.42.95.0 k8sx
- onap dev-drools-3 1/1 Running 1 1h 10.42.224.52 k8sx
- onap dev-drools-4 1/1 Running 0 1m 10.42.237.251 k8sx
- ..
+ RESOURCES:
+ ==> v1/ConfigMap
+ NAME DATA AGE
+ dev-dmaap-bus-controller-config 1 2d
+ dev-message-router-cadi-prop-configmap 1 2d
+ dev-message-router-msgrtrapi-prop-configmap 1 2d
+ dev-msb-discovery 1 2d
+ dev-msb-eag 1 2d
+ dev-msb-iag 1 2d
+ dev-brmsgw-pe-configmap 2 2d
+ dev-drools-configmap 6 2d
+ dev-drools-log-configmap 1 2d
+ dev-drools-settings-configmap 1 2d
+ dev-policydb-configmap 1 2d
+ dev-pdp-pe-configmap 3 2d
+ dev-pdp-log-configmap 1 2d
+ dev-pe-scripts-configmap 1 2d
+ dev-filebeat-configmap 1 2d
+ dev-pe-configmap 1 2d
+ dev-pap-pe-configmap 7 2d
+ dev-pap-log-configmap 1 2d
+ dev-pap-sdk-log-configmap 1 2d
+ dev-robot-resources-configmap 3 2d
+ dev-robot-lighttpd-authorization-configmap 1 2d
+ dev-robot-eteshare-configmap 4 2d
+
+ ==> v1/PersistentVolume
+ NAME CAPACITY ACCESS MODES RECLAIM POLICY STATUS CLAIM STORAGECLASS REASON AGE
+ dev-dbc-pg-data0 1Gi RWO Retain Bound onap/dev-dbc-pg-data-dev-dbc-pg-0 dev-dbc-pg-data 2d
+ dev-dbc-pg-data1 1Gi RWO Retain Bound onap/dev-dbc-pg-data-dev-dbc-pg-1 dev-dbc-pg-data 2d
+ dev-message-router-kafka 2Gi RWX Retain Bound onap/dev-message-router-kafka 2d
+ dev-message-router-zookeeper 2Gi RWX Retain Bound onap/dev-message-router-zookeeper 2d
+ dev-nexus 2Gi RWX Retain Bound onap/dev-nexus 2d
+ dev-policydb 2Gi RWX Retain Bound onap/dev-policydb 2d
+
+ ==> v1/PersistentVolumeClaim
+ NAME STATUS VOLUME CAPACITY ACCESS MODES STORAGECLASS AGE
+ dev-message-router-kafka Bound dev-message-router-kafka 2Gi RWX 2d
+ dev-message-router-zookeeper Bound dev-message-router-zookeeper 2Gi RWX 2d
+ dev-nexus Bound dev-nexus 2Gi RWX 2d
+ dev-policydb Bound dev-policydb 2Gi RWX 2d
+
+ ==> v1beta1/ClusterRoleBinding
+ NAME AGE
+ onap-binding 2d
+
+ ==> v1beta1/Deployment
+ NAME DESIRED CURRENT UP-TO-DATE AVAILABLE AGE
+ dev-dmaap-bus-controller 1 1 1 1 2d
+ dev-message-router-kafka 1 1 1 1 2d
+ dev-message-router-zookeeper 1 1 1 1 2d
+ dev-message-router 1 1 1 1 2d
+ dev-kube2msb 1 1 1 1 2d
+ dev-msb-consul 1 1 1 1 2d
+ dev-msb-discovery 1 1 1 1 2d
+ dev-msb-eag 1 1 1 1 2d
+ dev-msb-iag 1 1 1 1 2d
+ dev-brmsgw 1 1 1 1 2d
+ dev-nexus 1 1 1 1 2d
+ dev-policydb 1 1 1 1 2d
+ dev-pap 1 1 1 1 2d
+ dev-robot 1 1 1 1 2d
+
+ ==> v1/Pod(related)
+ NAME READY STATUS RESTARTS AGE
+ dev-dmaap-bus-controller-5bd859c7dc-blzdc 1/1 Running 0 2d
+ dev-message-router-kafka-748cdf7b9c-srv7l 1/1 Running 0 2d
+ dev-message-router-zookeeper-5b5969f6f-8rk9w 1/1 Running 0 2d
+ dev-message-router-b5bdc599c-5h56k 1/1 Running 0 2d
+ dev-kube2msb-579fc77c54-m84qx 1/1 Running 0 2d
+ dev-msb-consul-7bc4fcc8-94gsc 1/1 Running 0 2d
+ dev-msb-discovery-768547bcb-2hr7j 2/2 Running 0 2d
+ dev-msb-eag-5d95686c67-9lkzs 2/2 Running 0 2d
+ dev-msb-iag-675b649848-pv2gh 2/2 Running 0 2d
+ dev-brmsgw-5675f5877b-wv68s 1/1 Running 0 2d
+ dev-nexus-7d96568f5f-m8c4l 1/1 Running 0 2d
+ dev-policydb-587d55bdff-9gdjv 1/1 Running 0 2d
+ dev-pap-678b44cd87-wxbww 2/2 Running 0 2d
+ dev-robot-589c76bb6b-hrrdn 1/1 Running 0 2d
+ dev-dbc-pg-0 1/1 Running 0 2d
+ dev-dbc-pg-1 1/1 Running 0 2d
+ dev-drools-0 1/1 Running 1 2d
+ dev-drools-1 0/1 Init:0/1 0 1s
+ dev-pdp-0 2/2 Running 1 2d
+ dev-pdp-1 2/2 Running 0 33m
+
+ ==> v1/Secret
+ NAME TYPE DATA AGE
+ dev-dbc-pg Opaque 3 2d
+ dev-message-router-secret Opaque 1 2d
+ dev-drools-secret Opaque 2 2d
+ dev-policydb-secret Opaque 2 2d
+ onap-docker-registry-key kubernetes.io/dockercfg 1 2d
+
+ ==> v1beta1/StatefulSet
+ NAME DESIRED CURRENT AGE
+ dev-dbc-pg 2 2 2d
+ dev-drools 2 2 2d
+ dev-pdp 2 2 2d
+
+ ==> v1/Service
+ NAME TYPE CLUSTER-IP EXTERNAL-IP PORT(S) AGE
+ dbc-postgres ClusterIP 10.43.181.134 <none> 5432/TCP 2d
+ dbc-pg-replica ClusterIP 10.43.202.168 <none> 5432/TCP 2d
+ dbc-pg-primary ClusterIP 10.43.29.226 <none> 5432/TCP 2d
+ dmaap-bc NodePort 10.43.254.230 <none> 8080:30241/TCP,8443:30242/TCP 2d
+ message-router-kafka ClusterIP 10.43.69.159 <none> 9092/TCP 2d
+ message-router-zookeeper ClusterIP None <none> 2181/TCP 2d
+ message-router NodePort 10.43.123.102 <none> 3904:30227/TCP,3905:30226/TCP 2d
+ msb-consul NodePort 10.43.27.77 <none> 8500:30285/TCP 2d
+ msb-discovery NodePort 10.43.178.20 <none> 10081:30281/TCP 2d
+ msb-eag NodePort 10.43.77.235 <none> 80:30282/TCP,443:30284/TCP 2d
+ msb-iag NodePort 10.43.221.196 <none> 80:30280/TCP,443:30283/TCP 2d
+ brmsgw NodePort 10.43.21.222 <none> 9989:30216/TCP 2d
+ nexus NodePort 10.43.159.27 <none> 8081:30236/TCP 2d
+ drools NodePort 10.43.233.67 <none> 6969:30217/TCP,9696:30221/TCP 2d
+ policydb ClusterIP None <none> 3306/TCP 2d
+ pdp ClusterIP None <none> 8081/TCP 2d
+ pap NodePort 10.43.110.50 <none> 8443:30219/TCP,9091:30218/TCP 2d
+ robot NodePort 10.43.172.248 <none> 88:30209/TCP 2d
-
+Verify that the new PDP-D comes up with the latest policy coordinates:
+.. code-block:: bash
+ :caption: Verify new PDP-D 2 comes up with policies loaded
+
+ ubuntu@k8sx:~$ curl --silent --user '@1b3rt:31nst31n' -X GET http://10.42.172.88:9696/policy/pdp/engine/controllers/amsterdam/drools | python -m json.tool
+ {
+ "alive": true,
+ "artifactId": "policy-amsterdam-rules",
+ "brained": true,
+ "groupId": "org.onap.policy-engine.drools.amsterdam",
+ "locked": false,
+ "modelClassLoaderHash": 1657760388,
+ "recentSinkEvents": [],
+ "recentSourceEvents": [],
+ "sessionCoordinates": [
+ "org.onap.policy-engine.drools.amsterdam:policy-amsterdam-rules:0.5.0:closedloop-amsterdam"
+ ],
+ "sessions": [
+ "closedloop-amsterdam"
+ ],
+ "version": "0.5.0"
+ }
+
End of Document
-.. SSNote: Wiki page ref. https://wiki.onap.org/display/DW/Scalability%2C+Resiliency+and+Manageability
+.. SSNote: Wiki page ref. https://wiki.onap.org/display/DW/Policy+on+OOM
+.. SSNote: Old Wiki page ref. https://wiki.onap.org/display/DW/Scalability%2C+Resiliency+and+Manageability