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
When one make dependency between 2 DaemonSets we could detect that,and AppController could consider situation on each node separately rather than whole DaemonSet as one.
That would speedup deployment and it would be more robust.
The text was updated successfully, but these errors were encountered:
We will want to implement advanced DaemonSet pods orchestration on top of taint/tolerance/node affinity mechanisms as part of AppController pod.
One TODO is to make a proxy for DaemonSet status checker which will retrieve DaemonSet state from k8s storage and manipulate nodes/DS objects to orchestrate them properly. We will be able to orchestrate DS pods spawning by using taints and tolerances (e.g. if the parent DS is already running on this node, use proper taint to make child DS pod spawn there too).
Once this kubernetes/kubernetes#31136 will be merged user will be able to define it using regular kube api. But almost same behavior can be achieved now with deployments using affinity rules for order and anti-affinity for one copy per node.
When one make dependency between 2 DaemonSets we could detect that,and AppController could consider situation on each node separately rather than whole DaemonSet as one.
That would speedup deployment and it would be more robust.
The text was updated successfully, but these errors were encountered: