-
Notifications
You must be signed in to change notification settings - Fork 622
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Remove topic partitions #476
Comments
@rohitsalem is looking for reviews on ros2/rmw_fastrtps#192 until the OpenSplice and Connext pr's are opened. |
Not sure if there is some work by @wjwwood or @Karsten1987 to that effect, but the design document about ROS to DDS name mapping article will also need to be updated accordingly |
@rohitsalem Can you please embed the status of your latest CI here? |
Looks like this has been the victim of a new failure we have on windows recently: ros2/build_farmer#111 |
@ros2/team this PRs are ready for review, Thanks! |
So there are some (seemingly flaky) test failures between fastrtps and opensplice on windows. as we are not running Opensplice continuously on our farm, we could technically merge it. I'll have a look at the error messages as soon as I can. |
I can reproduce the behavior locally on my OSX machine as well when compiled in debug (I guess this is why the buildfarm didn't notice it). Only one combination fails: FastRTPS subscribes, Opensplice sends. Only for message type |
update: When increasing the publishing frequency from 10 Hz to 1 Hz the messages are getting transmitted correctly. Ways for me to reproduce the behavior: 1.) Start subscriber with FastRTPS on topic When applying 2.b it seems like OpenSplice so-to-say gives up and the message eventually goes through.
After debugging a fair amount of variations with @dhood, the tests will constantly pass when putting a sleep on the publisher side (OpenSplice) between advertising the publisher and starting the actual publishing process. That is, we can make sure that the publisher got correctly matched on fastrtps side. When publishing with a higher frequency, it seems fastrtps gets confused and can't handle the amount of incoming package while not having setup the subscriber correctly. So it looks to me as if they are two issues coming together here. The first one not being able to send messages if the publisher and subscriber aren't correctly set up and thus somewhat synchronized. I don't know how to fix that neither am I sure that this is really the case, but it simply looks like OpenSplice sets things up way too fast for FastRTPS to handle. It is correct to say though that having the API to wait for publisher and subscribers to have matched is a way to get these tests to pass. |
The issue is to track the removal of partitions usage in topics:
PR for rmw_fastrtps : ros2/rmw_fastrtps#192
PR for rmw_opensplice : ros2/rmw_opensplice#225
PR for rmw_connext : ros2/rmw_connext#285
PR for design doc : ros2/design#170
The text was updated successfully, but these errors were encountered: