NOTE: THIS DOCUMENT IS UNDER CONSTRUCTION AND NOT COMPLETED OR VETTED.
Written with StackEdit.
Written by Dale Bracey
Expert Resource - Matt Heler
Expert Resource - David Alfano
And understanding how it replicates data is hard.
“How much free space do I leave to the CRUSH map for fault-tolerance on node failures?”
“How many nodes can I lose and not lose data?”
The following aims to help “over-simplify” the answers to those questions. In this example, we use a 5-node Ceph environment. The data distribution is also a simplified representation that simply explains the number of copies across nodes of each object; but, does not show it in a manner that depicts a large number of files/objects stored on the nodes.
Ceph also has a customizable …
In this environment…
Pools = 1
Placement Groups (PG) = 1
OSDs (Disks) = ??
Replicas = 3
Objects = 5 (which just happen to all be of equal size and about a 1/4 the size of each node’s total available space)
Find more details at
https://ceph.com/geen-categorie/how-data-is-stored-in-ceph-cluster/
Here, you will see the 5-node Ceph environment with its
CEPH with 5 Nodes (Healthy State)
CEPH with 5 Nodes (1 Node Failed)
CEPH with 5 Nodes (2 Nodes Failed)
