Persistent Disk Storage and ARO
Google Compute Engine supports "zonal" and "regional" options for persistent disk storage that can be used for DR.
Figure: Storage scope types.
R1Z1/R1Z2 and R2Z1/R2Z2 are synchronous replications inside the same region used for HA. These disks are called "regional." Zonal disks do not have a secondary HA zone. R1Z1/R2Z1 is asynchronous replication used for DR applicable to both regional and zonal disks.
ARO does not manage synchronous replication failures. After a zonal failure, it is typically desired to keep VMs in the same region. Recovery after zonal failure should be conducted according to the procedure described in the article: Build HA services using synchronously replicated disks. Protection orchestration can then be resumed by mapping the newly created VMs onto the failed ones. At this point, asynchronous replication will be able to continue.
For example, if disk R1Z1 fails due to partial or complete zonal failure, a replica VM can be created using disk R1Z2. It can then be mapped through the AROVA UI. (Example: Regional VM, Primary Region, Primary Disk Primary Zone (R1Z1) Failure.)
ARO conducts failover if:
- The entire primary region R1 has failed.
- If a VM has a zonal disk in R1Z1 and the disk fails due to partial or full zonal failure.
- If a VM has a regional disk and both copies in R1Z1 and R1Z2 have failed.
- If at least a single dependent VM fails and must failover, the entire group of dependent VMs will failover together with it.
Also see:
View: Definitions, Acronyms and Abbreviations
View: Google Compute Engine and JetStream DR