You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
See #4265 (comment) for details
But TL;DR is that there are occassions when we don't want to do extra computing for every node update then coupling that together with say the fact that node updates for two different nodes in two different threads can happen in parallel that try to create the same ovn constructs can also cause race issues. In the past we have solved this using (1) waitops hack 2) mutexes 3) OVNK specific caches like what @trozet ^ is suggesting above 4) ensuring we call the create from a global space and avoid the threads racing all together if possible
I went with 2 for now, but probably 3 is the better way to do specially if in the future we have more cases like we do in case of node events.
The text was updated successfully, but these errors were encountered:
What is this issue about?
See #4265 (comment) for details
But TL;DR is that there are occassions when we don't want to do extra computing for every node update then coupling that together with say the fact that node updates for two different nodes in two different threads can happen in parallel that try to create the same ovn constructs can also cause race issues. In the past we have solved this using (1) waitops hack 2) mutexes 3) OVNK specific caches like what @trozet ^ is suggesting above 4) ensuring we call the create from a global space and avoid the threads racing all together if possible
I went with 2 for now, but probably 3 is the better way to do specially if in the future we have more cases like we do in case of node events.
The text was updated successfully, but these errors were encountered: