Delphix allows data objects to be replicated between Delphix Engines. Prior to version 5.3.3, these Delphix Engines had to be running identical Delphix versions, but could otherwise vary in terms of other configurations. With version 5.3.3 and above, engines can now be on different versions, which is further detailed in the Forward Compatible Replication (FCR) section. In the event of a failure that destroys the source engine, you can bring up the target engine in a state that matches the source. In addition, you can provision VDBs from replicated objects, allowing for the geographical distribution of data and remote provisioning.
You can run replication ad hoc, but it is typically run according to a predefined schedule. After the initial update, each subsequent update sends only the changes incurred since the previous update. Replication does not provide synchronous semantics, which would otherwise guarantee that all data is preserved on the target engine. When there is a failover to a replication target, some data is lost, equivalent to the last time a replication update was sent.
Replication is generally not suited for high-availability configurations where rapid failover (and failback) is a requirement. Failing over a replication target requires a non-trivial amount of time and is a one-way operation; failback requires replicating all data back to the original source. For cases where high availability is necessary, it is best to leverage features of the underlying hypervisor or storage platform. For more information on how to evaluate the use of Delphix Engine replication for your data recovery requirements, see the topics under Backup and Recovery Strategies for the Delphix Engine.
Forward compatible replication (FCR)
Delphix Virtualization supports the ability to replicate to a Delphix Target Engine running on a higher version. To do so, FCR has a few requirements and restrictions to consider:
FCR is supported for replication jobs starting from a source engine running 22.214.171.124 and beyond.
The target engine must be running 126.96.36.199 and beyond.
As of 188.8.131.52, the FCR replication operation can not go beyond engine versions released more than 12 months apart.
FCR supports replicating between major versions as long as those specific versions are no more than a year apart.
Examples of supported and not supported FCR configurations:
184.108.40.206 to 220.127.116.11 (same version)
18.104.22.168 to 22.214.171.124 (same version)
126.96.36.199 to 188.8.131.52 (higher patch version)
184.108.40.206 to 220.127.116.11
18.104.22.168 to 22.214.171.124
126.96.36.199 to 188.8.131.52
184.108.40.206 to 220.127.116.11 (source version not compatible with FCR)
18.104.22.168 to 22.214.171.124 (target version not compatible with FCR)
126.96.36.199 to 188.8.131.52 (source version higher than the target)
184.108.40.206 to 220.127.116.11 (there is more than a year between release dates for those versions)
Some of the newer versions of 5.3.x are not compatible with the early versions of 6.0.x.0 (18.104.22.168 and 22.214.171.124).
126.96.36.199 and 188.8.131.52 are not compatible with 184.108.40.206
220.127.116.11 and 18.104.22.168 are not compatible with 22.214.171.124
126.96.36.199 is not compatible with 188.8.131.52
184.108.40.206 is not compatible with 220.127.116.11 and 18.104.22.168
The following table lists all the Delphix Engine versions that a user can replicate to the required or the latest version.
When replicating from X.Y.X.* to X.Y.Z.* the target version has to be greater than or equal to the source version. We are using the * convention to refer to patch releases and reduce the table size.
From Source Version
To Target Version
5.3.2.* - 6.0.9.*
5.3.3.* - 6.0.9.*
5.3.4.* - 6.0.9.*
5.3.5.* - 6.0.9.*
5.3.6.* - 6.0.9.*
5.3.7.* - 5.3.9.*
6.0.1.* - 6.0.9.*
5.3.8.* - 5.3.9.*
6.0.1.* - 6.0.9.*
6.0.2.* - 6.0.9.*
6.0.0.* - 6.0.7.*
6.0.1.* - 6.0.7.*
6.0.2.* - 6.0.7.*
6.0.3.* - 6.0.7.*
6.0.4.* - 6.0.7.*
6.0.5.* - 6.0.11.*
6.0.6.* - 6.0.12.*
6.0.7.* - 6.0.13.*
6.0.8.* - 6.0.14.*
6.0.9.* - 6.0.15.*
6.0.10.* - 6.0.16.*
6.0.11.* - 6.0.17.*
6.0.12.* - 7.0.0.*
6.0.13.* - 9.0.0.*
6.0.14.* - 11.0.0.*
6.0.15.* - 13.0.0.*
7.0.0.* - 16.0.0.*
8.0.0.* - 16.0.0.*
9.0.0.* - 16.0.0.*
10.0.0.* - 16.0.0.*
11.0.0.* - 16.0.0.*
12.0.0.* - 16.0.0.*
13.0.0.* - 16.0.0.*
14.0.0.* - 16.0.0.*
15.0.0.* - 16.0.0.*
As virtual appliances, it is possible to backup, restore, replicate, and migrate data objects between Delphix Engines using features of VMWare and the underlying storage infrastructure. Data objects include groups, dSources, VDBs, Self-Service (Jet Stream) data templates and data containers, and associated dependencies. In addition to the replication capabilities provided by this infrastructure, native Delphix Engine replication provides further capabilities, such as the ability to replicate a subset of objects, replicate multiple sources to a single target, and provision VDBs from replicated dSources and VDBs without affecting ongoing updates. The topics under Backup and Recovery Strategies for the Delphix Engine provide more information on how to evaluate features of the Delphix Engine in relation to your backup and recovery requirements.
Replication is configured on the source Delphix Engine and copies a subset of dSources and VDBs to a target Delphix Engine. It then sends incremental updates manually or according to a schedule. For more information on configuring replication, see Configuring Replication.
You can use replicated dSources and VDBs to provision new VDBs on the target side. You can refresh these VDBs to data sent as part of an incremental replication update, as long as you do not destroy the parent object on the replication source. For more information, see Provisioning from Replicated Data Sources or VDBs.
During replication, replicated dSources and VDBs are maintained in an alternate replica and are not active on the target side. In the event of a disaster, a failover operation can break the replication relationship. For more information on how to activate replicated objects, see Replicas and Failover.
When you select objects for replication, the engine will automatically include any dependencies, including parent objects, such as groups, and data dependencies such as VDB sources. This means that replicating a VDB will automatically include its group, the parent dSource, and the group of the dSource, as well as any environments associated with those databases. When replicating an entire engine, all environments will be included. When replicating a database or group, only those environments with the replicated databases are included.
Only database objects and their dependencies, as well as certain non-database objects, are copied as part of a backup or replication operation, including:
Self-service (Jet Stream) Data Templates and Data Containers
Environment configuration (users, database instances, and installations)
Delphix users, roles, permissions, and authorizations
Database configuration templates
The following objects are NOT copied as part of a backup or replication operation:
Events and faults
System services settings, such as SMTP
After failover, you must recreate these settings on the target.
On-Premises Replication To Azure/OCI/GCP/Hyper-V
Replicating from on-premises engines with an underlying storage block size of 512B will experience disk usage inflation when replicating to target engines with different underlying block sizes. Azure, GCP, Hyper-V, and OCI are known to have 4K block sizes and therefore will require extra disk capacity when receiving replication from an on-premises engine. This behavior is due to the fact that the underlying storage block size is different (512B vs. 4K) between the two Delphix Engines (one on Prem, one on Azure/OCI/GCP/Hyper-V), resulting in a lower compression rate on the replication target. It is expected that 1.5-1.6x the amount of space is taken from objects on-premises in these cases.
Resumable replication enhances the current replication feature by allowing you to restart large, time-consuming initial replication or incremental updates from an intermediate point. A single replication instance can fail for a number of environmental and internal reasons. Previously, when you restarted a failed replication instance, replication required a full resend of all data transmitted prior to the failure. With resumable replication, no data is retransmitted. Replication is resumable across machine reboot, stack restart, and network partitions. The resumable replication feature is fully automated and does not require or allow any user intervention.
For example, suppose a replication profile has already been configured from a source to a target. A large, full send begins between the two that is expected to take weeks to complete. Halfway through, a power outage at the data center that houses the source causes the source machine to go down and only come back up after a few hours. On startup, the source will detect a replication was ongoing, automatically re-contact the target, and resume the replication where it left off. In the user interface (UI) on the source, the same replication sends job will appear as active and continue to update its progress. However, in the UI of the target, a new replication receives job will appear but will track its progress as a percentage of the entire replication.
In 4.1 and earlier releases, the replication component would always clean up after failed jobs to ensure that the Delphix Engine was kept in a consistent state and that no storage was wasted on unused data. With the addition of reusability, the target and source can choose to retain a partial replication state following a failure to allow future replications to complete from that intermediate point. In the current release, the target and source will only choose to retain partial replication state following failures that leave them out of network contact with each other – for example, source restart, target restart, or network partition. Once network contact is re-established, the ongoing replication will be automatically detected and resumed.
Replication will not resume after failures that leave the source and target connected. For example, if a storage failure on the target, such as out-of-space errors, causes a replication to fail, the source and target remain connected. As a result, the Engine will discard state data associated with the failed Replication operation. Nonetheless, resumable replication would begin during a source reboot, a target reboot, and a network partition.
Replicating Delphix self-service templates
Templates can now be replicated and accessed on the target engine via Delphix Self-Service (Jet Stream). Replicated templates can be replicated into the target space with or without their containers. On the new target engine, the newly created replicated template can be used to create new containers that are assigned to users. You cannot change the replicated template’s name or the names of the containers from which it was replicated.
Any containers that were replicated over with the template cannot be used to start, stop, etc until they are disconnected to their parent containers in the source engine during the failover operation.
Replication of non-data objects
As of 22.214.171.124, the replication of non-data objects is supported.
Non-data objects refer to:
Delphix users, roles, permissions, and authorizations
Database configuration templates
These objects will not be presented as selectable objects when creating a replication spec, but will instead be passively included by association, the same way environments are. Replication of non-data objects will follow these rules:
If the entire engine is replicated, all non-data objects will be included
If specific data objects are replicated, all associated non-data objects will be included
For example, when a container is replicated, then policies that apply to that container, as well as users who have authorizations on that container, will all be included for replication.
Replicated users will be shown on the Received Replicas page on the Delphix Engine UI.
Also as of 126.96.36.199, the automatic conflict resolution option is chosen by default; it is the recommended way of handling the failover of non-data objects, as non-data objects are expected to cause collisions. Automatic conflict resolution will resolve non-data object conflicts by the following rules:
Users will be consolidated if both username and email match. Otherwise, the replicated user will be renamed.
Roles, permissions, and authorizations will be consolidated
Policies and database configuration templates will be renamed, except for "None" type policies, which will be consolidated. Following failover, replicated policies will continue to apply to the same replicated data objects.