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 you run a collaborative cluster, you may not know how many nodes there are at any given time. But you like to add a min/max replication value, to not spread all pins to all nodes - if there are many.
Say I want to save 2 TB of 200 pins. If there are 500 nodes it would be nice to have just 10-15 copies of every pin.
So each node just have to save a fraction of the whole pinset.
But when I start out, I don't run 10 nodes. This means I can't set it to 10-15 but need to set it to 1-15. But this means, the pins may get lost, as the cluster will only act on not having the minimum redundancies.
The solution would be, to set the replication level to opportunistic 10-15. So the cluster would put up to 15 members into the allocation list, but won't block the pinning if it's below 10.
When a new cluster member comes online, the cluster would reevaluate which pins are below the minimum redundancy and add the new member to the allocation list.
The text was updated successfully, but these errors were encountered:
Describe the feature you are proposing
When you run a collaborative cluster, you may not know how many nodes there are at any given time. But you like to add a min/max replication value, to not spread all pins to all nodes - if there are many.
Say I want to save 2 TB of 200 pins. If there are 500 nodes it would be nice to have just 10-15 copies of every pin.
So each node just have to save a fraction of the whole pinset.
But when I start out, I don't run 10 nodes. This means I can't set it to 10-15 but need to set it to 1-15. But this means, the pins may get lost, as the cluster will only act on not having the minimum redundancies.
The solution would be, to set the replication level to opportunistic 10-15. So the cluster would put up to 15 members into the allocation list, but won't block the pinning if it's below 10.
When a new cluster member comes online, the cluster would reevaluate which pins are below the minimum redundancy and add the new member to the allocation list.
The text was updated successfully, but these errors were encountered: