When physical disk failures occur on multiple nodes, the primary concern is the impact on data availability. The effects of such failures vary depending on the data redundancy policy.
If the redundancy policy is replication, as long as one replica remains available when multiple physical disks across different nodes fail, data can still be read and written normally, and cluster services will not be affected. However, if the failures render all replicas unavailable and unrecoverable, data will be completely lost, which will affect services.
If the data redundancy policy is erasure coding, under normal conditions, data is stored as K data blocks and M parity blocks across different nodes. When physical disks on multiple nodes fail, as long as no fewer than K blocks (data or parity blocks) remain available, normal read and write operations can be maintained, and the cluster services will not be affected. If fewer than K coded blocks remain, the data will become inaccessible, which will affect the normal operation of cluster services.
