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
Introduce a new option to limits.cpu.nodes to allow for balancing between NUMA nodes.
This would affect both containers and virtual machines and would basically look at all NUMA nodes on the system on instance startup, pick whichever currently has the lowest usage, store that in a volatile key and then use it as the actual limits.cpu.nodes value within the rest of our logic.
We can refine the logic to identify the least used NUMA node down the line.
Initially we can keep this pretty simple by just mapping out how many instances CPUs are allocated to each node. We'd treat any case where limits.cpu is empty to mean 1 CPU for simplicity.
The text was updated successfully, but these errors were encountered:
Introduce a new option to
limits.cpu.nodes
to allow for balancing between NUMA nodes.This would affect both containers and virtual machines and would basically look at all NUMA nodes on the system on instance startup, pick whichever currently has the lowest usage, store that in a volatile key and then use it as the actual
limits.cpu.nodes
value within the rest of our logic.We can refine the logic to identify the least used NUMA node down the line.
Initially we can keep this pretty simple by just mapping out how many instances CPUs are allocated to each node. We'd treat any case where
limits.cpu
is empty to mean 1 CPU for simplicity.The text was updated successfully, but these errors were encountered: