CDP Replication Chain
A replication chain is a sequence of files that allows you to roll back a replica to a specific point in time during failover. Veeam Backup & Replication creates replication chains for each VM added to a CDP policy.
The replication chain contains short-term and long-term restore points. Short-term restore points allow you to roll back replica data to a state created seconds or minutes ago, while long-term restore points — to a state created hours or days ago. Short-term restore points are always crash-consistent, long-term restore points can be crash-consistent or application-consistent depending on how long-term restore points are configured in CDP policy settings.
The replication chain is stored on the target datastore in the <replica_VM_name> folder. The replication chain consists of the files of the following types (only the key file types are listed):
- VMX — the configuration file of the replica.
The replication chain contains one .vmx file, other files from the list are created per virtual disk.
- VMDK — virtual disk files that store contents of replica hard disk drives.
On the datastore, you can see files under the following names:
- <disk_name>.vmdk — files that store full copies of virtual disks, that is, store base disk data. These files are created during the initial synchronization and relate to the very first long-term restore point. This restore point is crash-consistent.
- <disk_name>-<index>.vmdk — files that store incremental changes made to virtual disks, that is, store delta disk data. Delta disk files relate to the following:
- Long-term restore points. In this case, files are created according to the long-term schedule configured in CDP policy settings. These files remain unchanged till they are retained by long-term retention.
- Technical points. These technical points are required during short-term retention. The technical points creation is connected to the transaction log creation. The technical point is created when the transaction log file reaches its maximum size or when transaction log is created after the short-term retention is reached. For more information on transaction log files, see the TLOG description.
- <disk_name>-<index>.tlog.vmdk — transaction log files that store incremental changes made to virtual disks during RPO set in CDP policy settings. These files relate to short-term restore points that are created once in RPO set in CDP policy settings. One transaction log file stores data for multiple short-term restore points.
New transaction log files are created in the following cases:
- When a transaction log file reaches a specific size: 2 TB on VMFS datastores, 512 GB on VSAN and VVOL.
- When a long-term restore point is created.
- When the short-term retention is reached. For example, if you set short-term retention to 1 hour, a new transaction log file will be created every hour.
- After failback commit.
- <disk_name>-interim.vmdk — files for protective virtual disks. Changes made to virtual disks will be written in these files when you perform failover.
Although VMDK files look like VMware snapshot files, they are not real snapshots. These files are created by I/O filter installed on the target host.
- VMFD — files that contains metadata for disks.
- META — files that contain metadata for transaction log files.
To roll back a replica to a specific point in time, the chain of files created for the replica must contain files with data for the base disk (<disk_name>-flat.vmdk) and a set of files that contain incremental changes for disks (<disk-name>-<index>.vmdk + .tlog.vmdk). If any file in the replication chain is missing, you will not be able to roll back to the necessary state. For this reason, you must not delete files from the datastore manually. Instead, you must specify retention policy settings that will let you maintain the desired number of files. For more information on retention policies and how to configure them, see Retention Policies.