Loki with multiple instances


I’m trying to understand how Loki managed local indexes before pushing them to the object store in a distributed environment.
The recommended way is to use boltdb shipper but from what I understand, there is a delay before the indexes are sent to the long term storage. If the container or pod is destroyed before sending, the index is lost.
I’m trying to setup Loki on ECS.

What i’m missing ?

Thank you!

If you are concerned about single point of failure on the write path then you should look into using replication factor of more than 1.