Mixing best effort and reliable topics in the same node causes best effort topics to drop data.

[cwit-vcas]

Generated by Generative AI

No

Operating System:

Linux ros2 6.14.0-35-generic #35~24.04.1-Ubuntu SMP PREEMPT_DYNAMIC Tue Oct 14 13:55:17 UTC 2 x86_64 x86_64 x86_64 GNU/Linux

ROS version or commit hash:

jazzy

RMW implementation (if applicable):

rmw_zenoh_cpp

RMW Configuration (if applicable):

No response

Client library (if applicable):

rclcpp

'ros2 doctor --report' output

No response

Steps to reproduce issue

1. Create a talker node with a best effort and a reliable QoS publisher.
2. Create a listener node with a best effort and reliable QoS subscriber to these topics.

I have attached an MWE here.

qos_issues_mwe.zip

Expected behavior

That both callbacks are processed equally at the same rate in the listener node.

Actual behavior

For publishing at 10 Hz I see the best effort topic dropping more than 9/10 messages. Very few messages get through, but notably some does make it.

If you run the MWE you will see messages like

[INFO] [1762433029.494035672] [qos_listener]: [best effort + volatile] Receive rate: 0.00 msg/s over 1.00 s (total=0) [INFO] [1762433029.494412061] [qos_listener]: [reliable + volatile] Receive rate: 9.99 msg/s over 1.00 s (total=10) [INFO] [1762433030.493976739] [qos_listener]: [best effort + volatile] Receive rate: 1.00 msg/s over 1.00 s (total=1) [INFO] [1762433030.494135978] [qos_listener]: [reliable + volatile] Receive rate: 10.00 msg/s over 1.00 s (total=20)

Showing that few messages get through on the best effort topic.
Changing the best effort publisher to reliable will make all messages get through at full rate.
Same goes for changing the reliable publisher to best effort.

Additional information

No response

[cwit-vcas]

This seems like a quite standard use case to me.

[cwit-vcas]

It works on rmw_fastrtps_cpp so I belive it is a Zenoh bug.

[sloretz]

@cwit-vcas thank you for the MRE!

Do I understand correctly that whether or not the best_effort() listener on the /topic_best_effort gets messages depends on the QoS settings of the publisher on an unrelated topic /topic_reliable?

Changing the best effort publisher to reliable will make all messages get through at full rate. Same goes for changing the reliable publisher to best effort.

This one is especially surprising to me. Best Effort publishers are not supposed to be compatible with reliable subscribers, so I would expect changing the publisher on /topic_reliable to best_effort() would cause the subscriber on the /topic_reliable to receive no messages. (See QoS Compatibility table).

🧇 @JEnoch Would you be willing to try reproducing this using the MRE above?

[cwit-vcas]

@sloretz
Yes that is correct. But as far as I understand (looking at the source code) with Zenoh best effort and reliable topics are compatible in all combinations?

In the meantime we have tested with a source release and it seems fixed in the upstream, so it was only on the ubuntu jazzy packages the issue was present.
I believe it was this pull request which fixed the issue: eclipse-zenoh/zenoh#2152?

So I think this issue can be closed.

[cwit-vcas]

Hopefully the upstream changes made it to the jazzy sync yesterday, but haven't had time to test that yet.

[JEnoch]

@cwit-vcas Thanks for your MWE.I confirm I reproduce the issue on jazzy with rmw_zenoh 0.2.8, but not 0.2.9.
The issue was fixed in Zenoh 1.6.2, which is now shipped with:

• rmw_zenoh 0.1.8 for humble (not yet synced)
• rmw_zenoh 0.2.9 for jazzy
• rmw_zenoh 0.6.6 for kilted
• rmw_zenoh 0.10.2 for rolling (not yet synced)

Please confirm you no longer see the issue with latest jazzy sync.