Disk failure - how many disks can fail before data loss?

Disclaimer: I do not work for EMC and this is not an official response, but I just saw this question and wanted to help.

Best Case (with replica): the entire cluster can fail and your application will transparently continue to read content from the replica without interruption.

Best Case (without replica): Disks can continue to fail without data loss if there is sufficient remaining capacity within the cluster to hold the data and there is enough time (~ 24 hours) between the failure events for the failed disk's data to be regenerated onto spare capacity within the cluster

Worst Case (without replica): A second disk failure on a different node within a cube (8 nodes) before the first failure has been regenerated can potentially lose data if the two disks host CPP fragments of the same c-clip.

Regards,

Mike Horgan

4 node cluster with Replica - how many disks can fail without data loss in ATMOS gen 2