As part of the Red Hat UKI Professional Services team, I have worked with several customers who are implementing AMQ Broker on Red Hat OpenShift Container Platform (OCP). One question customers typically ask is, “How do we validate that the AMQ configuration is correct for our scenario?” Previously, I would have suggested one of the following:
These tools can give you indicators around:
- Is the broker up and running? That is, can it receive/publish messages for this configuration?
- Can the broker handle a certain performance characteristic? That is, what is my minimum publish rate per second for this configuration?
- And much more.
The problem with these tools is that you cannot choose the client technology. This could lead to real-world differences and limited technology choices, which in turn might lead you down the wrong technology path. In other words:
- Do you get the same performance from JMeter versus the AMQ clients you would use in production? Are you comparing like for like? Apples with apples?
So, what do I think is the answer? Quiver [1]. In this article, I'll provide an overview and demo of using Quiver with Red Hat AMQ on Red Hat OpenShift. If you're looking for more information on Red Hat AMQ and how it can help, check out this webinar.
What is Quiver?
Straight from the Quiver GitHub page:
Quiver implementations are native clients (and sometimes also servers) in various languages and APIs that either send or receive messages and write raw information about the transfers to standard output. They are deliberately simple.
The quiver-arrow command runs a single implementation in send or receive mode and captures its output. It has options for defining the execution parameters, selecting the implementation, and reporting statistics.
The quiver command launches a pair of quiver-arrow instances, one sender and one receiver, and produces a summary of the end-to-end transmission of messages.
In my own words;
Quiver is a tool that starts both a publisher and receiver client with the aim of completing as fast as possible. Once the publisher and receiver have completed; a statistical output is generated which can give you insight into how the test has performed.
The clients are based on the Apache Qpid project that offers; Java, C++, Python and Javascript implementations to name a few. The clients primarily target JMS and AMQP users.
Recorded demo
Feeling lazy and want to watch an asciinema recording?
The demo will show the following:
- AMQ Broker and AMQ Interconnect being deployed onto OKD running on Minishift
- Send and receive messages via Quiver using the Core protocol to the broker
- Send and receive messages via Quiver using a JMS client over AMQP protocol to the broker
- Send and receive messages via Quiver using a JMS client over AMQP protocol to interconnect
Do it yourself demo
Want to play with Quiver, AMQ, and OCP yourself?
The demo uses an OCP template which requires two arguments:
- DOCKER_IMAGE; location of the fully qualified docker pull URL. This can be resolved via the imported image stream;
oc get is quiver -o jsonpath=‘{.status.dockerImageRepository}’ - DOCKER_CMD; the quiver command, in JSON array format, you want to execute.
Demo prerequisites
It is presumed that you have the following:
- Basic knowledge of the OCP CLI and web console
- A running Red Hat OpenShift Container Platform (OCP) or Minishift environment
- AMQ Broker templates and image streams installed by following this documentation: 2.1. Installing AMQ Broker on OpenShift Container Platform image streams and application templates
- AMQ Interconnect templates and image streams installed by following: 2.1. Verifying the availability of AMQ Interconnect templates
Setup
- Create project
$ oc new-project quiver
- Import the Quiver image
$ oc import-image quiver:latest --from=docker.io/ssorj/quiver --confirm
- Add the view role to the default service account
$ oc policy add-role-to-user view system:serviceaccount:$(oc project -q):default
- Deploy AMQ Broker
$ oc new-app --template=amq-broker-72-basic \ -e AMQ_PROTOCOL=openwire,amqp,stomp,mqtt \ -e AMQ_QUEUES=quiver \ -e AMQ_ADDRESSES=quiver \ -e AMQ_USER=anonymous \ -e ADMIN_PASSWORD=password $ oc get pods --watch - Deploy AMQ Interconnect
$ oc new-app --template=amq-interconnect-1-basic $ oc get pods --watch
Send messages to AMQ Broker
- Send 1,000,000 messages to the broker via the core protocol
$ oc process -f https://raw.githubusercontent.com/ssorj/quiver/0.2.0/packaging/openshift/openshift-pod-template.yml \ DOCKER_IMAGE=$(oc get is quiver -o jsonpath='{.status.dockerImageRepository}'):latest \ DOCKER_CMD="[\"quiver\", \"//broker-amq-tcp:61616/activemq-artemis-jms\", \"--impl\", \"activemq-artemis-jms\", \"--verbose\"]" \ | oc create -f - - Look at output for core protocol pod
$ oc get pods --watch $ oc logs -f {insert value of new pod name} ---------------------- Sender ----------------------- --------------------- Receiver ---------------------- -------- Time [s] Count [m] Rate [m/s] CPU [%] RSS [M] Time [s] Count [m] Rate [m/s] CPU [%] RSS [M] Lat [ms] ----------------------------------------------------- ----------------------------------------------------- -------- 2.3 0 0 73 64.6 2.3 0 0 73 63.8 0 4.3 3,337 1,667 71 100.0 4.3 2,529 1,261 75 96.6 42 6.3 24,727 10,684 71 116.0 6.4 25,178 11,223 106 123.1 36 8.3 113,170 44,177 64 116.7 8.4 116,495 45,658 66 123.9 2 10.3 221,942 54,359 59 116.8 10.4 227,809 55,491 50 124.0 2 12.3 327,983 52,994 56 116.8 12.4 332,016 52,077 44 122.3 2 14.3 428,866 50,391 54 116.8 14.4 433,091 50,512 41 122.5 2 16.3 540,823 55,923 62 117.1 16.4 547,056 56,926 49 122.6 2 18.3 645,499 52,312 60 117.1 18.4 648,975 50,934 48 122.6 2 20.3 761,704 58,073 58 117.1 20.4 768,000 59,453 55 122.7 2 22.3 873,509 55,875 61 117.2 22.4 874,857 53,348 50 122.7 3 24.3 914,014 20,242 33 117.2 24.5 923,768 22,728 27 122.7 4 26.3 1,000,000 42,950 53 117.4 26.5 1,000,000 38,078 45 0.0 2 -------------------------------------------------------------------------------- Subject: activemq-artemis-jms //broker-amq-tcp:61616/activemq-artemis-jms (/tmp/quiver-xo6l0q4a) Count: 1,000,000 messages Body size: 100 bytes Credit window: 1,000 messages Duration: 23.0 s Sender rate: 43,414 messages/s Receiver rate: 43,507 messages/s End-to-end rate: 43,407 messages/s Latencies by percentile: 0%: 0 ms 90.00%: 4 ms 25%: 2 ms 99.00%: 20 ms 50%: 2 ms 99.90%: 249 ms 100%: 307 ms 99.99%: 290 ms -------------------------------------------------------------------------------- - Send 1,000,000 messages to the broker via the JMS client over AMQP protocol
$ oc process -f https://raw.githubusercontent.com/ssorj/quiver/0.2.0/packaging/openshift/openshift-pod-template.yml \ DOCKER_IMAGE=$(oc get is quiver -o jsonpath='{.status.dockerImageRepository}'):latest \ DOCKER_CMD="[\"quiver\", \"amqp://broker-amq-amqp:5672/qpid-jms-over-ampq\", \"--impl\", \"qpid-jms\", \"--verbose\"]" \ | oc create -f - - Look at output for the AMQP client pod
$ oc get pods --watch $ oc logs -f {insert value of new pod name} ---------------------- Sender ----------------------- --------------------- Receiver ---------------------- -------- Time [s] Count [m] Rate [m/s] CPU [%] RSS [M] Time [s] Count [m] Rate [m/s] CPU [%] RSS [M] Lat [ms] ----------------------------------------------------- ----------------------------------------------------- -------- 2.2 2,663 1,330 109 103.1 2.2 1,984 991 94 94.6 89 4.2 16,304 6,817 118 110.7 4.2 16,113 7,050 115 113.2 56 6.2 45,720 14,693 102 119.7 6.2 45,809 14,767 97 115.4 118 8.2 105,097 29,674 88 120.8 8.2 92,810 23,489 84 115.6 305 10.2 159,018 26,920 71 120.8 10.2 146,602 26,883 76 115.6 411 12.2 213,794 27,361 81 121.5 12.2 205,564 29,466 81 117.2 400 14.2 271,138 28,643 77 121.6 14.2 256,132 25,271 76 117.2 461 16.2 325,058 26,960 82 116.6 16.2 311,757 27,785 75 117.2 566 18.2 383,747 29,315 78 116.6 18.2 362,224 25,208 72 117.2 678 20.2 431,554 23,880 64 116.6 20.2 411,578 24,665 70 117.2 750 22.2 492,444 30,430 75 116.6 22.2 468,113 28,225 77 117.2 786 24.2 537,684 22,609 70 116.7 24.2 510,287 21,087 70 117.3 1,105 26.2 595,639 28,963 77 116.7 26.2 567,833 28,759 75 117.4 1,049 28.2 646,014 25,175 67 116.7 28.2 619,412 25,790 73 117.5 1,014 30.2 701,401 27,680 73 116.9 30.2 669,677 25,095 72 117.5 1,167 32.2 757,523 28,019 73 116.9 32.2 722,166 26,231 72 117.6 1,246 34.2 803,007 22,731 70 117.1 34.2 768,587 23,210 66 117.6 1,360 36.2 856,193 26,593 69 117.1 36.2 816,627 23,924 73 117.6 1,538 38.2 911,459 27,605 73 117.1 38.2 870,633 26,990 77 117.6 1,633 40.2 968,803 28,658 70 117.4 40.2 926,359 27,849 76 117.6 1,482 ---------------------- Sender ----------------------- --------------------- Receiver ---------------------- -------- Time [s] Count [m] Rate [m/s] CPU [%] RSS [M] Time [s] Count [m] Rate [m/s] CPU [%] RSS [M] Lat [ms] ----------------------------------------------------- ----------------------------------------------------- -------- 42.2 1,000,000 15,583 41 0.0 42.2 978,444 26,029 89 117.6 1,581 - - - - - 44.2 1,000,000 10,773 57 0.0 1,608 -------------------------------------------------------------------------------- Subject: qpid-jms amqp://broker-amq-amqp:5672/qpid-jms-over-ampq (/tmp/quiver-a1xz3nbz) Count: 1,000,000 messages Body size: 100 bytes Credit window: 1,000 messages Duration: 42.4 s Sender rate: 24,696 messages/s Receiver rate: 23,616 messages/s End-to-end rate: 23,573 messages/s Latencies by percentile: 0%: 1 ms 90.00%: 1575 ms 25%: 534 ms 99.00%: 1741 ms 50%: 988 ms 99.90%: 1886 ms 100%: 1929 ms 99.99%: 1926 ms --------------------------------------------------------------------------------
Send messages to AMQ Interconnect
- Send 1,000,000 messages to the interconnect via the JMS client over AMQP protocol
$ oc process -f https://raw.githubusercontent.com/ssorj/quiver/0.2.0/packaging/openshift/openshift-pod-template.yml \ DOCKER_IMAGE=$(oc get is quiver -o jsonpath='{.status.dockerImageRepository}'):latest \ DOCKER_CMD="[\"quiver\", \"amqp://amq-interconnect:5672/qpid-jms-over-ampq-via-interconnect\", \"--impl\", \"qpid-jms\", \"--verbose\"]" \ | oc create -f - - Look at output for AMQP client pod
$ oc get pods --watch $ oc logs -f {insert value of new pod name} ---------------------- Sender ----------------------- --------------------- Receiver ---------------------- -------- Time [s] Count [m] Rate [m/s] CPU [%] RSS [M] Time [s] Count [m] Rate [m/s] CPU [%] RSS [M] Lat [ms] ----------------------------------------------------- ----------------------------------------------------- -------- 2.2 1,529 764 131 102.6 2.2 1,465 731 120 84.5 21 4.2 12,580 5,514 151 116.2 4.2 12,529 5,526 138 114.5 16 6.2 42,121 14,756 125 119.5 6.2 41,815 14,606 90 111.3 6 8.2 85,563 21,721 82 119.9 8.2 85,028 21,596 71 111.7 4 10.2 125,149 19,773 75 120.2 10.2 124,858 19,905 62 112.0 5 12.2 169,899 22,264 79 116.5 12.2 169,965 22,486 67 112.1 5 14.3 212,816 21,437 69 116.7 14.2 212,442 21,217 60 112.1 5 16.3 259,889 23,536 75 116.7 16.2 259,571 23,553 71 114.7 4 18.3 308,674 24,332 86 119.1 18.2 308,723 24,539 69 114.9 4 20.3 340,342 15,826 67 119.1 20.2 340,176 15,726 58 115.0 6 22.3 386,804 23,208 76 119.3 22.2 386,597 23,199 67 115.0 4 24.3 434,611 23,880 75 119.3 24.2 434,435 23,895 65 115.0 4 26.3 472,270 18,783 63 119.4 26.2 471,754 18,650 53 115.0 6 28.3 507,116 17,406 59 119.4 28.2 506,646 17,420 51 115.0 6 30.3 541,841 17,345 57 119.4 30.2 541,639 17,488 56 115.1 6 32.3 577,421 17,781 61 119.4 32.2 576,429 17,352 50 115.1 6 34.3 612,512 17,528 61 119.4 34.2 612,231 17,892 53 115.1 6 36.3 644,913 16,192 58 119.5 36.2 644,190 15,972 48 115.1 6 38.3 678,782 16,901 58 119.5 38.2 678,577 17,176 52 115.1 7 40.3 714,240 17,720 60 119.5 40.2 713,569 17,487 51 115.3 6 ---------------------- Sender ----------------------- --------------------- Receiver ---------------------- -------- Time [s] Count [m] Rate [m/s] CPU [%] RSS [M] Time [s] Count [m] Rate [m/s] CPU [%] RSS [M] Lat [ms] ----------------------------------------------------- ----------------------------------------------------- -------- 42.3 750,187 17,965 62 119.5 42.2 748,967 17,672 51 115.3 6 44.3 787,234 18,496 63 119.5 44.3 787,500 19,171 55 115.3 6 46.3 820,124 16,420 58 119.5 46.3 819,155 15,820 48 115.3 6 48.3 856,316 18,087 59 119.5 48.3 855,868 18,338 51 115.3 6 50.3 897,520 20,551 68 119.5 50.3 897,232 20,641 57 115.3 5 52.3 944,594 23,537 73 119.5 52.3 944,159 23,452 63 115.3 4 54.3 990,567 22,986 72 119.5 54.3 990,176 22,997 64 115.3 4 56.3 1,000,000 4,716 17 0.0 56.3 1,000,000 4,907 15 0.0 4 -------------------------------------------------------------------------------- Subject: qpid-jms amqp://amq-interconnect:5672/qpid-jms-over-ampq-via-interconnect (/tmp/quiver-0sttx8_z) Count: 1,000,000 messages Body size: 100 bytes Credit window: 1,000 messages Duration: 53.7 s Sender rate: 18,615 messages/s Receiver rate: 18,634 messages/s End-to-end rate: 18,614 messages/s Latencies by percentile: 0%: 0 ms 90.00%: 10 ms 25%: 3 ms 99.00%: 18 ms 50%: 5 ms 99.90%: 43 ms 100%: 89 ms 99.99%: 72 ms --------------------------------------------------------------------------------
Cleanup
- Cleanup and delete all quiver pods
$ oc delete pod -l app=quiver
Does your client use another language?
If your language of choice was not shown in the demo, the good news is that Quiver supports several implementations. Here are some examples:
$ oc process -f https://raw.githubusercontent.com/ssorj/quiver/0.2.0/packaging/openshift/openshift-pod-template.yml \
DOCKER_IMAGE=$(oc get is quiver -o jsonpath='{.status.dockerImageRepository}'):latest \
DOCKER_CMD="[\"quiver\", \"//broker-amq-tcp:61616/activemq-jms\", \"--impl\", \"activemq-jms\", \"--verbose\"]" \
| oc create -f -
$ oc process -f https://raw.githubusercontent.com/ssorj/quiver/0.2.0/packaging/openshift/openshift-pod-template.yml \
DOCKER_IMAGE=$(oc get is quiver -o jsonpath='{.status.dockerImageRepository}'):latest \
DOCKER_CMD="[\"quiver\", \"//broker-amq-tcp:61616/qpid-messaging-cpp\", \"--impl\", \"qpid-messaging-cpp\", \"--verbose\"]" \
| oc create -f -
$ oc process -f https://raw.githubusercontent.com/ssorj/quiver/0.2.0/packaging/openshift/openshift-pod-template.yml \
DOCKER_IMAGE=$(oc get is quiver -o jsonpath='{.status.dockerImageRepository}'):latest \
DOCKER_CMD="[\"quiver\", \"//broker-amq-tcp:61616/qpid-messaging-python\", \"--impl\", \"qpid-messaging-python\", \"--verbose\"]" \
| oc create -f -
$ oc process -f https://raw.githubusercontent.com/ssorj/quiver/0.2.0/packaging/openshift/openshift-pod-template.yml \
DOCKER_IMAGE=$(oc get is quiver -o jsonpath='{.status.dockerImageRepository}'):latest \
DOCKER_CMD="[\"quiver\", \"//broker-amq-tcp:61616/qpid-proton-c\", \"--impl\", \"qpid-proton-c\", \"--verbose\"]" \
| oc create -f -
$ oc process -f https://raw.githubusercontent.com/ssorj/quiver/0.2.0/packaging/openshift/openshift-pod-template.yml \
DOCKER_IMAGE=$(oc get is quiver -o jsonpath='{.status.dockerImageRepository}'):latest \
DOCKER_CMD="[\"quiver\", \"//broker-amq-tcp:61616/qpid-proton-cpp\", \"--impl\", \"qpid-proton-cpp\", \"--verbose\"]" \
| oc create -f -
$ oc process -f https://raw.githubusercontent.com/ssorj/quiver/0.2.0/packaging/openshift/openshift-pod-template.yml \
DOCKER_IMAGE=$(oc get is quiver -o jsonpath='{.status.dockerImageRepository}'):latest \
DOCKER_CMD="[\"quiver\", \"//broker-amq-tcp:61616/qpid-proton-python\", \"--impl\", \"qpid-proton-python\", \"--verbose\"]" \
| oc create -f -
$ oc process -f https://raw.githubusercontent.com/ssorj/quiver/0.2.0/packaging/openshift/openshift-pod-template.yml \
DOCKER_IMAGE=$(oc get is quiver -o jsonpath='{.status.dockerImageRepository}'):latest \
DOCKER_CMD="[\"quiver\", \"//broker-amq-tcp:61616/rhea\", \"--impl\", \"rhea\", \"--verbose\"]" \
| oc create -f -
Food for thought
I hope this demo has shown how easy it is to use Quiver to interact with AMQ Broker and AMQ Interconnect on OCP.
It should have also raised questions around how Quiver could be integrated into your own systems. For example:
- Could Quiver be part of your mvn integration tests?
- Could Quiver be added as a stage within your CI/CD pipeline for smoke testing?
- Could the results of the smoke test pass/fail the deployment?
- Could Quiver be used as part of a heartbeat monitoring system to validate that the broker is alive?
These are several scenarios for which I see Quiver being highly useful, and you’ve probably thought of others.
Note
[1] Although Quiver is developed by Red Hat employees, it is not supported under a Red Hat subscription and is strictly an upstream project.
Last updated: November 14, 2023