Open headlee opened 12 months ago
There is a humble patch release pending to update Fast DDS to v2.6.6, which has several improvements related to SHM and resilience to crashing nodes.
@headlee We've now released Humble Patch release 5: https://github.com/ros2/ros2/releases/tag/release-humble-20230925 . Can you check to see if the problem you reported is fixed? Thanks.
The new patch release has definitely helped. We still see some intermittent issues but it's unclear at this point if they are due to fastrtps, ros daemon on Windows, or something else.
Is there any downside to us deleting the contents of the fastrtps_interprocess folder (C:\ProgramData\eprosima\fastrtps_interprocess
) before every launch of our app? That has fixed issues in the past for us.
Is there any downside to us deleting the contents of the fastrtps_interprocess folder (
C:\ProgramData\eprosima\fastrtps_interprocess
) before every launch of our app? That has fixed issues in the past for us.
@headlee No downsides. You could still do that.
Is there any downside to us deleting the contents of the fastrtps_interprocess folder (
C:\ProgramData\eprosima\fastrtps_interprocess
) before every launch of our app? That has fixed issues in the past for us.@headlee No downsides. You could still do that.
Great, thanks!
One more sort of related question: do you know if it's possible to change the directory that eprosima uses to store the shm files on Windows? I found datasharingqospolicy in the docs but I don't know if that's for normal shm or just the DataReader/DataWriter delivery. Haven't had a chance to test it out yet.
do you know if it's possible to change the directory that eprosima uses to store the shm files on Windows?
@headlee I'm afraid it cannot be changed. It is fixed here
Bug report
Required Info:
Operating System: Windows 10
Installation type: Binaries
Version or commit hash: Humble (Patch Release 4, 43083e5)
DDS implementation: Fast-RTPS
Client library (if applicable): rclcpp
Steps to reproduce issue
Sometimes when we run our launchfile one or more of our nodes will immediately die or seemingly hang. It seems to be caused by using Fast-RTPS shared memory on Windows...we are able to get around some of these issues by doing
fastdds.bat shm clean
but today we got the error pasted below and aren't sure how to proceed.If anyone has any advice to make this reproducible and/or knows how to fix it that would be great. We're hoping that a new Windows Humble patch release with Fast-RTPS/Fast-DDS version bumps (which should solve this) could also help us out, but any advice in the meantime is appreciated!
Additional information
Error message:
Fast-RTPS profile: