Support dynamic concurrency control #220
sebastianburckhardt
announced in
Announcements
Replies: 1 comment
|
Looks like the problem disappears when setting maxConcurrentActivityFunctions = 10. So the problem seems to be that CPU overload (100 cpu-intensive tasks running at same time) wreaks havoc on the lease management, causing terrible churn. Since I am not seeing the same issue for AzureStorage backend, I guess what is happening is that the latter "self-regulates": if the system is overloaded, it cannot fetch more activities to execute. Here however, the activities are all local and ready to go. I believe we need some regulation mechanism so that if the node is already struggling with CPU load, it does not start new activities, but lets them sit in the queue from where they can be redistributed. |
0 replies
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Uh oh!
There was an error while loading. Please reload this page.
Looks like the Locavore load balancing is not working on the latest bits.
Specifically, this test:
is not showing any signs of load rebalancing, all activities are executed locally.
All reactions