Bug #818854 address one possible cause of an upload scheduling meltdown. However, if a backend is down or otherwise unreachable for a sufficient period of time, there will still be a significant backlog of scheduled uploads; even though these will most likely not be scheduled for exactly the same time, there will likely be a significant number of uploads running concurrently, which could cause sufficient resource starvation to achieve meltdown, and is most likely undesirable even when it does not lead to a complete meltdown.
The correct approach for handling this is probably to control concurrency independently for each backend.
Imported from Launchpad using lp2gh.
Bug #818854 address one possible cause of an upload scheduling meltdown. However, if a backend is down or otherwise unreachable for a sufficient period of time, there will still be a significant backlog of scheduled uploads; even though these will most likely not be scheduled for exactly the same time, there will likely be a significant number of uploads running concurrently, which could cause sufficient resource starvation to achieve meltdown, and is most likely undesirable even when it does not lead to a complete meltdown.
The correct approach for handling this is probably to control concurrency independently for each backend.
Imported from Launchpad using lp2gh.