Interesting longhorn replica eviction behavior I bumped into. So:
• disabled scheduling on all disks on a longhorn node
• Set eviction to true on the disks on the same node
As there’s no other node to actually “take on” the evicted replicas nothing happens.
I enable disk scheduling again. However, without setting eviction requested to false on all the disks on the node.
The replicas on the node gets doubled.
I know I may be fooling around a bit here. But, still - isn’t this unintended behavior?