Simple example of running pbench agent in a container

contrib/containerized-pbench/.gitignore

@@ -0,0 +1,3 @@
+fiotest
+*.tmp
+var_lib_pbench-agent

contrib/containerized-pbench/README.md

@@ -1,124 +1,10 @@
-## Containerized pbench
-This directory contains OpenShift templates, dockerfiles needed to build and run containerized pbench.
-
-### Prepare the nodes you want to monitor
-The nodes need to have pbench-agent images on it. In case the nodes do not have the images, please pull them from dockerhub:
-
-#### pbench-agent
-```
-$ docker pull ravielluri/image:agent
-$ docker tag ravielluri/image:agent pbench-agent:latest
-```
-
-### Label the nodes with a type=pbench label
-```   
-$ oc label node <node> type=pbench
-```
-
-### Create a pbench namespace
-```   
-$ oc create -f pbench/contrib/containerized-pbench/openshift-templates/pbench-namespace.yml
-```
-
-### Create a service account and add it to the privileged Security Context Constraints
-```
-$ oc create serviceaccount useroot
-$ oc adm policy add-scc-to-user privileged -z useroot
-```
-
-### Create pbench-agent pods and patch the daemonset
-```
-$ oc create -f pbench/contrib/containerized-pbench/openshift-templates/pbench-agent-daemonset.yml
-$ oc patch daemonset pbench-agent --patch \ '{"spec":{"template":{"spec":{"serviceAccountName": "useroot"}}}}'
-```
-
-## Prepare the jump host to run pbench-controller
-
-### Get the controller image from dockerhub
-```
-$ docker pull ravielluri/image:controller
-$ docker tag ravielluri/image:controller pbench-controller:latest
-```
-
-### keys
-copy the ssh keys to /root/scale-testing/keys. The keys directory should contain a perf key named as id_rsa_perf,  id_rsa - the private key needed to copy the results to the pbench server, authorized_keys file -ansible needs to have a passwordless authentication to localhost inside the container, ssh config which looks like:
-```
-Host *
-	User root
-        Port 2022
-        StrictHostKeyChecking no
-        PasswordAuthentication no
-        UserKnownHostsFile ~/.ssh/known_hosts
-        IdentityFile ~/.ssh/id_rsa_perf  
-
-Host *pbench-server
-        User root
-        Port 22
-        StrictHostKeyChecking no
-        PasswordAuthentication no
-        UserKnownHostsFile ~/.ssh/known_hosts
-        IdentityFile /opt/pbench-agent/id_rsa
-```
-
-### Inventory
-Make sure you have the inventory used to install openshift and it should look like:
-```
-[pbench-controller]
-
-[masters]
-
-[nodes]
-
-[etcd]
-
-[lb]
-
-[prometheus-metrics]
-<host> port=8443 cert=/etc/origin/master/admin.crt key=/etc/origin/master/admin.key
-<host> port=10250 cert=/etc/origin/master/admin.crt key=/etc/origin/master/admin.key
-
-[pbench-controller:vars]
-register_all_nodes=False
-```
-
-Set register_all_nodes to true if the tools needs to be registered on all the nodes, if not set to true, it registers pbench tools on just two of the nodes.
-
-NOTE: 
-- Make sure all the variables are defined under [group:vars], all the stuff under [groups] are assumed to be the node ip’s.
-
-- In HA environment, we will have an lb which is not an openshift node. This means that there won’t be a pbench-agent pod running on the lb, pbench-ansible will fail registering tools as it won’t find a pbench-agent pod. So, we need to make a copy of the original inventory and get rid of the lb node from the inventory which is being mounted into the container.
-
-- We need to make sure we stick to either ip’s or hostnames in both inventory and openshift for certificates to be valid.
-
-## Run benchmarks
-Edit the vars file and set benchmark_type variable to the benchmark that you want to run, pbench_server - host where the results are moved when move_results is set to True.
-
-### Avalaible benchmark_type options:
-- nodeVertical
-- http
-- masterVertical
-
-## Run
-```
-$ docker_id=$(docker run -t -d --name=controller --net=host --privileged \
-  -v /var/lib/pbench-agent:/var/lib/pbench-agent \
-  -v <path to results directory>:/var/lib/pbench-agent \
-  -v <path to the inventory>:/root/inventory \
-  -v <path to the vars file>:/root/vars \
-  -v <path to the keys>:/root/.ssh \
-  -v <path to benchmark.sh>:/root/benchmark.sh pbench-controller
-```
-
-### Results
-If move_results is set, the results are moved to the pbench server. In case you want to look at the results before moving, set move_results to False and the results will be available in the mounted directory.
-
-### Monitoring the benchmark
-In case you want to monitor what's going on inside the container, please take a look at the logs:
-
-```
-$ docker logs -f pbench-controller
-```
-You should be able to see the pbench-ansible registering the tools and the benchmark stdout.
-
-### Need for privileged container
-Docker needs to be run under privileged mode to get access to the host system's devices. Host's /var/lib/pbench-agent directory, /proc are mounted on to the container.
+# Running a Pbench Agent-driven Workload in a Container
+1. Create a workload script (`example-workload.sh` can be used to start)
+   This script performs the invocation of a Pbench Agent workload driver, e.g.
+   `pbench-user-benchmark`, or `pbench-fio`, or `pbench-uperf`.
+2. Setup the local execution environment to be aware of the target Pbench Server
+   `./setup.sh <pbench server host name>:<port number>`
+3. Execute the workload
+   Typically this script performs the necessary setup required for the
+   containerized environment (`example-driver.sh` can be used to start, note
+   that it also moves the results to the Pbench Server).

contrib/containerized-pbench/example-driver.sh

@@ -0,0 +1,45 @@
+#!/bin/bash
+
+if [[ ! -f ./var_lib_pbench-agent/.token || ! -f ./pbench-agent.cfg.tmp ]]; then
+    echo "Please run ./setup.sh to properly setup the required environment." >&2
+    exit 1
+fi
+
+# Create a `fiotest` directory to use as the target directory of the `fio`
+# workload execution.
+mkdir -p fiotest
+
+# Run the workload in the container.
+#
+# The `example-workload.sh` script is mounted into the container as
+# `/workload.sh`, along with the target `fiotest` directory it requires, and the
+# location of `/var/lib/pbench-agent`.  This example relies on the CentOS 8
+# Pbench Agent "all" container image (which has everything and the kitchen sink
+# installed).
+#
+# Notice that host networking is used for the container.
+#
+# NOTE that the container is not run with any privileges or adjusted namespaces
+# so the tool data collected may not be what you expect.
+podman run \
+    --rm \
+    --network host \
+    --name example-workload \
+    -v ${HOME}/.ssh:/root/.ssh:z \
+    -v ./example-workload.sh:/workload.sh:z \
+    -v ./var_lib_pbench-agent:/var/lib/pbench-agent:z \
+    -v ./fiotest:/fiotest:z \
+    quay.io/pbench/pbench-agent-all-centos-8:latest /workload.sh
+
+# Move the results to the target Pbench Server
+#
+# Note that we run the mover separately to demonstrate that the collected data
+# is independent of the containerized environment in which it was collected.
+podman run \
+    -it \
+    --rm \
+    --network host \
+    -v ./mover.sh:/mover.sh:z \
+    -v ./pbench-agent.cfg.tmp:/opt/pbench-agent/config/pbench-agent.cfg:z \
+    -v ./var_lib_pbench-agent:/var/lib/pbench-agent:z \
+    quay.io/pbench/pbench-agent-all-centos-8:latest /mover.sh

contrib/containerized-pbench/example-workload.sh

@@ -0,0 +1,19 @@
+#!/bin/bash
+# An example of how to run a simple workload with `pbench-user-benchmark`
+
+# In a containerized environment, the Pbench Agent profile must be run to
+# establish the proper execution environment.
+source /opt/pbench-agent/profile
+
+# The set of tools the user desires to run must be registered before the
+# workload is executed.
+echo "Registering the 'heavy' tool set provided by the Pbench Agent"
+pbench-register-tool-set heavy
+
+# The `user-benchmark` workload wrapper simply takes a configuration name, in
+# this case the very generic `example-workload`, followed by the command to use
+# to run the workload.  Note that the command is separated by `--` to avoid
+# any workload command line arguments that follow from being interpreted as
+# options to `pbench-user-benchmark` command itself.
+echo "Executing pbench-user-benchmark ..."
+pbench-user-benchmark --config example-workload -- fio --directory=/fiotest --name fio_test_file --direct=1 --rw=randread --bs=16k --size=100M --numjobs=16 --time_based --runtime=20s --group_reporting --norandommap

contrib/containerized-pbench/gen-token.sh

@@ -0,0 +1,9 @@
+#!/bin/bash
+
+# In a containerized environment, the Pbench Agent profile must be run to
+# establish the proper execution environment.
+source /opt/pbench-agent/profile
+
+# Generate a token to move results to the target Pbench Server storing it in the
+# mounted /var/lib/pbench-agent directory.
+pbench-generate-token --output=/var/lib/pbench-agent/.token