This overview describes classic protection mechanisms. Deployment-specific recovery instructions take precedence, and these mechanisms do not replace tested backups. The article is not a failure-injection or recovery runbook.
Part of the Teradata Internals and Operations guide.
Teradata offers a range of features to safeguard against potential data loss, each providing distinct and granular protection. While some features are enabled by default, others require explicit activation.
1. The Teradata Transient Journal – Transaction Failure Protection
Transaction changes are logged in the Teradata transient journal, also known as the transaction log.
Teradata tables are typically distributed across all AMPs based on their Primary Index values, but this is not the case for the transient journal. Instead, each AMP maintains its own transient journal.
The transient journal preserves the original row, maintaining data integrity in case of row modification. This permits the system to revert the table to its initial state in the event of an error. The rows are only temporarily held in the transient journal and are deleted after successful transaction completion. The parsing engine guarantees that all AMPs finish their tasks successfully before committing any transactions.
Get the complete Teradata tuning book, free
Teradata Query Performance Tuning: The DWHPro Guide
By Roland Wenzlofsky & Artemios Vogiatzis
206 pages on statistics and demographics, access paths and indexes, query execution plans, and joins. Get the full PDF and practical Teradata updates from DWHPro.
Confirm your email to receive your download link. Privacy policy. Unsubscribe in any newsletter.
About the free book2. Teradata Fallback – AMP Failure Protection
The transient journal safeguards your data solely from transaction failures. However, in the event of an AMP failure, your data may become inaccessible until the AMP is restored. The fallback protection strategy was introduced to guarantee the accessibility of data in such circumstances.
Each row is stored twice for fallback protection, with a secondary hash map designating a fallback AMP for each row and preserving a duplicate. Tables with fallback protection require double the space.
Fallback protection allows for data accessibility in the event of an AMP failure because the auxiliary AMP will perform its tasks until the failed AMP is restored.
Availability depends on the actual failed components and copy placement. For example, the documented VMware fallback-rebuild procedure requires no more than one failed AMP per fallback cluster; this is not a guarantee of recovery from arbitrary multiple-AMP failures.
Fallback placement uses configured AMP clusters; a cluster is not necessarily a fixed pair of AMPs. Failure tolerance depends on where the primary and fallback copies reside.
The primary and fallback AMPs are not stored together on the same physical node, which is a wise hardware design decision to prevent failures. If a node completely fails, the fallback protection allows access to the data.
3. RAID 1 – Disk Failure Protection
Fallback protection safeguards against failure of the Unix processes (namely, VPROC and AMP) in the Teradata System.
RAID, virtual disks and storage redundancy are deployment-specific. The RAID 1 explanation below describes a mirrored-storage configuration, not a requirement that every current Teradata deployment use that layout.
Like with fallback protection, using this feature incurs double the disk space usage.
Implementing a RAID 1 configuration can optimize data accessibility and security. In the event of a disk failure, the mirror disk would serve as a backup, preventing any data loss. However, data may be permanently lost if both the primary and mirror disks fail and no fallback protection is in place.
4. The Teradata Clique – Node Failure Protection
The Clique configuration provides an additional layer of security to the Teradata System by protecting against node failures. Group nodes, called Cliques, are formed to cluster AMPs for improved security.
The AMPs from the failed node will be transferred to another node within the same clique and remain fully operational. However, the recipient node will incur a heavier workload, leading to a detrimental effect on performance.
Where supported and configured, a hot standby node can take over work from a failed node. Recovery still has operational and performance effects; standby capacity does not guarantee zero disruption or unchanged latency.
Restart and recovery steps depend on the platform, release and failure scenario. Follow the supported recovery procedure for that environment rather than applying a universal reboot sequence.
Deciding what to change in your data platform?
Compare optimisation, modernisation and migration against your workload and business requirements.
Get the complete Teradata tuning book, free
Teradata Query Performance Tuning: The DWHPro Guide
By Roland Wenzlofsky & Artemios Vogiatzis
206 pages on statistics and demographics, access paths and indexes, query execution plans, and joins. Get the full PDF and practical Teradata updates from DWHPro.
Confirm your email to receive your download link. Privacy policy. Unsubscribe in any newsletter.
About the free book
Hi Roland,
Thanks for the awesome post.
Is it right that a permanent journal too protects the data and can make up for the 5th way for protecting data?
Why it does not make the cut?
Great Post ! Thanks a lot !!
awesome explanation!!