Adding support of P-UDN using VMs

[capolrik]

Based on #158

Needs to be run on OCP 4.18 and above.

[capolrik]

I was able to generate some results using the l2bridge binding method in OCP 4.18 but passt method is not stable for this version. I keep this PR as a draft until this feature becomes stable.

[capolrik]

I have added an option for switching from l2bridge and passt binding method. The l2bridge has still the best performance on 4.18.rc5 but passt performance and stability is improving.

[capolrik]

@jtaleric is it ok to merge this one?

[smalleni]

@jtaleric is it ok to merge this one?

@capolrik hey, got a link to the results from running this?

[jtaleric]

@capolrik can you rebase? sorry!

[capolrik]

@jtaleric do you have the time for a little review?

[capolrik]

@jtaleric @smalleni @venkataanil, someone has few minutes to review this?

[josecastillolema]

Thanks for adding the documentation @capolrik

[capolrik]

here is an example of an output:

$ ./bin/amd64/k8s-netperf --udnl2 --vm --udnPluginBinding=l2bridge --vm-image=quay.io/openshift-cnv/qe-cnv-tests-fedora:36
INFO[2025-03-19 10:41:13] Starting k8s-netperf (cnv-udn@6a2fadde14a2cb64469c4eda590d873504ef8279)
INFO[2025-03-19 10:41:13] 📒 Reading netperf.yml file.
INFO[2025-03-19 10:41:13] 📒 Reading netperf.yml file - using ConfigV2 Method.
INFO[2025-03-19 10:41:13] Cleaning resources created by k8s-netperf
INFO[2025-03-19 10:41:13] ⏰ Waiting for netperf Namespace to be deleted...
INFO[2025-03-19 10:42:21] 🔬 prometheus discovered at openshift-monitoring
INFO[2025-03-19 10:42:21] 🔨 Creating namespace: netperf
INFO[2025-03-19 10:42:21] 🔨 Creating service account: netperf
INFO[2025-03-19 10:42:21] Deploying L2 Primary UDN in the NS : netperf
WARN[2025-03-19 10:42:21] ⚠️  No zone label
WARN[2025-03-19 10:42:21] ⚠️  Single node per zone and/or no zone labels
INFO[2025-03-19 10:42:21] 🚀 Creating service for iperf-service in namespace netperf
INFO[2025-03-19 10:42:21] 🚀 Creating service for uperf-service in namespace netperf
INFO[2025-03-19 10:42:22] 🚀 Creating service for netperf-service in namespace netperf
W0319 10:42:22.242683 1030729 warnings.go:70] metadata.finalizers: "foregroundDeleteVirtualMachine": prefer a domain-qualified finalizer name to avoid accidental conflicts with other finalizer writers
INFO[2025-03-19 10:42:22] 🚀 Creating service for client-across-svc in namespace netperf
INFO[2025-03-19 10:42:22] ⏰ Wating for VMI (client-across) to be in state running
INFO[2025-03-19 10:43:44] Looking for pods with label app=client-across
W0319 10:43:44.740242 1030729 warnings.go:70] metadata.finalizers: "foregroundDeleteVirtualMachine": prefer a domain-qualified finalizer name to avoid accidental conflicts with other finalizer writers
INFO[2025-03-19 10:43:44] ⏰ Wating for VMI (server) to be in state running
INFO[2025-03-19 10:44:09] Looking for pods with label app=server
INFO[2025-03-19 10:44:14] Connecting via ssh to the VMI
INFO[2025-03-19 10:44:14] 🗒️  Running netperf TCP_STREAM (service false) for 10s
INFO[2025-03-19 10:46:58] 🗒️  Running netperf TCP_STREAM (service false) for 10s
INFO[2025-03-19 10:47:11] 🗒️  Running netperf TCP_STREAM (service false) for 10s
INFO[2025-03-19 10:47:24] 🗒️  Running netperf TCP_STREAM (service false) for 10s
INFO[2025-03-19 10:47:37] 🗒️  Running netperf TCP_STREAM (service false) for 10s
INFO[2025-03-19 10:47:51] 🗒️  Running netperf TCP_STREAM (service false) for 10s
INFO[2025-03-19 10:48:04] 🗒️  Running netperf UDP_STREAM (service false) for 10s
INFO[2025-03-19 10:48:17] 🗒️  Running netperf UDP_STREAM (service false) for 10s
INFO[2025-03-19 10:48:31] 🗒️  Running netperf UDP_STREAM (service false) for 10s
INFO[2025-03-19 10:48:44] 🗒️  Running netperf TCP_CRR (service false) for 10s
INFO[2025-03-19 10:48:59] 🗒️  Running netperf TCP_CRR (service false) for 10s
INFO[2025-03-19 10:49:14] 🗒️  Running netperf TCP_CRR (service false) for 10s
INFO[2025-03-19 10:49:30] 🗒️  Running netperf TCP_RR (service false) for 10s
INFO[2025-03-19 10:49:43] 🗒️  Running netperf TCP_RR (service false) for 10s
INFO[2025-03-19 10:49:56] 🗒️  Running netperf TCP_RR (service false) for 10s
+-------------------+---------+------------+-------------+--------------+---------+-----------------+--------------+-------+-----------+----------+---------+--------------------+---------------------------------+
|    RESULT TYPE    | DRIVER  |  SCENARIO  | PARALLELISM | HOST NETWORK | SERVICE | EXTERNAL SERVER | MESSAGE SIZE | BURST | SAME NODE | DURATION | SAMPLES |     AVG VALUE      |     95% CONFIDENCE INTERVAL     |
+-------------------+---------+------------+-------------+--------------+---------+-----------------+--------------+-------+-----------+----------+---------+--------------------+---------------------------------+
| 📊 Stream Results | netperf | TCP_STREAM | 1           | false        | false   | false           | 1024         | 0     | false     | 10       | 3       | 4185.950000 (Mb/s) | 3678.497289-4693.402711 (Mb/s)  |
| 📊 Stream Results | netperf | TCP_STREAM | 1           | false        | false   | false           | 8192         | 0     | false     | 10       | 3       | 8717.226667 (Mb/s) | 6743.795535-10690.657799 (Mb/s) |
| 📊 Stream Results | netperf | UDP_STREAM | 1           | false        | false   | false           | 1024         | 0     | false     | 10       | 3       | 1662.896667 (Mb/s) | 1182.860093-2142.933241 (Mb/s)  |
+-------------------+---------+------------+-------------+--------------+---------+-----------------+--------------+-------+-----------+----------+---------+--------------------+---------------------------------+
+---------------+---------+----------+-------------+--------------+---------+-----------------+--------------+-------+-----------+----------+---------+---------------------+----------------------------------+
|  RESULT TYPE  | DRIVER  | SCENARIO | PARALLELISM | HOST NETWORK | SERVICE | EXTERNAL SERVER | MESSAGE SIZE | BURST | SAME NODE | DURATION | SAMPLES |      AVG VALUE      |     95% CONFIDENCE INTERVAL      |
+---------------+---------+----------+-------------+--------------+---------+-----------------+--------------+-------+-----------+----------+---------+---------------------+----------------------------------+
| 📊 Rr Results | netperf | TCP_CRR  | 1           | false        | false   | false           | 1024         | 0     | false     | 10       | 3       | 4066.773333 (OP/s)  | 3773.107469-4360.439198 (OP/s)   |
| 📊 Rr Results | netperf | TCP_RR   | 1           | false        | false   | false           | 1024         | 0     | false     | 10       | 3       | 10492.870000 (OP/s) | 10188.762502-10796.977498 (OP/s) |
+---------------+---------+----------+-------------+--------------+---------+-----------------+--------------+-------+-----------+----------+---------+---------------------+----------------------------------+
+--------------------+---------+----------+-------------+--------------+---------+-----------------+--------------+-------+-----------+----------+---------+-------------------+
|    RESULT TYPE     | DRIVER  | SCENARIO | PARALLELISM | HOST NETWORK | SERVICE | EXTERNAL SERVER | MESSAGE SIZE | BURST | SAME NODE | DURATION | SAMPLES | AVG 99%TILE VALUE |
+--------------------+---------+----------+-------------+--------------+---------+-----------------+--------------+-------+-----------+----------+---------+-------------------+
| RR Latency Results | netperf | TCP_CRR  | 1           | false        | false   | false           | 1024         | 0     | false     | 10       | 3       | 333.000000 (usec) |
| RR Latency Results | netperf | TCP_RR   | 1           | false        | false   | false           | 1024         | 0     | false     | 10       | 3       | 156.333333 (usec) |
+--------------------+---------+----------+-------------+--------------+---------+-----------------+--------------+-------+-----------+----------+---------+-------------------+
+---------------------+---------+------------+-------------+--------------+---------+-----------------+--------------+-------+-----------+----------+---------+-------------+
|        TYPE         | DRIVER  |  SCENARIO  | PARALLELISM | HOST NETWORK | SERVICE | EXTERNAL SERVER | MESSAGE SIZE | BURST | SAME NODE | DURATION | SAMPLES |  AVG VALUE  |
+---------------------+---------+------------+-------------+--------------+---------+-----------------+--------------+-------+-----------+----------+---------+-------------+
| TCP Retransmissions | netperf | TCP_STREAM | 1           | false        | false   | false           | 1024         | 0     | false     | 10       | 3       | 456.333333  |
| TCP Retransmissions | netperf | TCP_STREAM | 1           | false        | false   | false           | 8192         | 0     | false     | 10       | 3       | 2099.666667 |
| UDP Loss Percent    | netperf | UDP_STREAM | 1           | false        | false   | false           | 1024         | 0     | false     | 10       | 3       | 0.004336    |
+---------------------+---------+------------+-------------+--------------+---------+-----------------+--------------+-------+-----------+----------+---------+-------------+
INFO[2025-03-19 10:50:10] Cleaning resources created by k8s-netperf

[josecastillolema]

Why do you need this export?

[capolrik]

I added it when I was troubleshooting the cloudinit process but it should have been remove. Deleting it

[jtaleric]

I don't see is adding any metadata to let us know the binding in the results. Lets add that to the output of both the CSV and the JSON/ES document.

[capolrik]

@jtaleric I have added 'UDN info' which is the udn layer + Binding method as an output, it looks like this:

+---------------------+---------+------------+-------------+--------------+---------+-----------------+-------------------+--------------+-------+-----------+----------+---------+-------------+
|        TYPE         | DRIVER  |  SCENARIO  | PARALLELISM | HOST NETWORK | SERVICE | EXTERNAL SERVER |     UDN INFO      | MESSAGE SIZE | BURST | SAME NODE | DURATION | SAMPLES |  AVG VALUE  |
+---------------------+---------+------------+-------------+--------------+---------+-----------------+-------------------+--------------+-------+-----------+----------+---------+-------------+
| TCP Retransmissions | netperf | TCP_STREAM | 1           | false        | false   | false           | layer2 - l2bridge | 1024         | 0     | false     | 10       | 3       | 360.333333  |
| TCP Retransmissions | netperf | TCP_STREAM | 1           | false        | false   | false           | layer2 - l2bridge | 8192         | 0     | false     | 10       | 3       | 3313.000000 |
| UDP Loss Percent    | netperf | UDP_STREAM | 1           | false        | false   | false           | layer2 - l2bridge | 1024         | 0     | false     | 10       | 3       | 0.004719    |
+---------------------+---------+------------+-------------+--------------+---------+-----------------+-------------------+--------------+-------+-----------+----------+---------+-------------+

For pod UDN, it displays only the layer.