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
Is your feature request related to a problem? Please describe.
Description:
Zone is a common concept in Cloud and important for High Availability. When a zone-1 is down, we don't want to see a collector from zone-1 failing to scrape targets from zone-2 or other healthy zones and causing missing metric datapoints in healthy zones. Each zone should have a way to be isolated.
A per-zone allocation strategy implements an allocation strategy that only assigns targets from a specific zone to collectors from the same zone.
Describe the solution you'd like
My Solution is to develop a new implementation of Allocation Strategy called "per-zone". On a zone awareness enabled node (e.g. AWS EC2), there's a label "topology. kubernetes.io/zone" which can be utilized.
The text was updated successfully, but these errors were encountered:
I think rather than introducing a zone specific strategy it would make more sense for us to introduce a partitioned strategy which takes an argument for a partition key. This would allow a user to partition on any key not just zone or node.
Hey Jacob, that's really a good idea to have one single partition strategy while making the partition key configurable. I will spend some after-work hours to think about the implementation. Still not sure if this is an easy work.
Component(s)
target allocator
Is your feature request related to a problem? Please describe.
Description:
Zone is a common concept in Cloud and important for High Availability. When a zone-1 is down, we don't want to see a collector from zone-1 failing to scrape targets from zone-2 or other healthy zones and causing missing metric datapoints in healthy zones. Each zone should have a way to be isolated.
A per-zone allocation strategy implements an allocation strategy that only assigns targets from a specific zone to collectors from the same zone.
Describe the solution you'd like
My Solution is to develop a new implementation of Allocation Strategy called "per-zone". On a zone awareness enabled node (e.g. AWS EC2), there's a label "topology. kubernetes.io/zone" which can be utilized.
The text was updated successfully, but these errors were encountered: