Closed dgraeber closed 8 months ago
@a13zen and @kevinsoucy are the points of reference. @dgraeber has tested the removal of lines 63-67 and with 0 nodes created, the ASG did indeed create a new instance and was not blocked with k8s.io/cluster-autoscaler/node-template/taint/dedicated
removed.
as per @a13zen this is ok to remove
The ASG taint in ml-training/k8s-managed/configure_asgs.py (line 63-67 that set
k8s.io/cluster-autoscaler/node-template/taint/dedicated
taint) was considered needed, but preliminary testing indicates it blocks GPU instances from scaling up from 0 nodes.Need someone to confirm that it CAN BE removed from the codebase