TB113 Continuous Data Engine May Become Unresponsive During Migration to Elastic Data

Impact

Continuous Data Engines migrating to leverage Elastic Data may become non-responsive. Elastic Data, a separately-licensed feature available for Continuous Data Engines, allows the use of various object storage backends (e.g. Amazon AWS S3, Azure Blob, etc.) instead of traditional block storage.  In the DevOps Platform 15.0 release, Delphix added a feature that allows for the in-place migration to object storage of an engine originally configured with block storage.  Upon the successful completion of a migration operation, the originally configured block storage can be reclaimed. 

As a result of the issue described in this bulletin, an affected Continuous Data Engine may become persistently non-responsive, and migrations may not complete normally. Because migration operations are irreversible and non-cancellable, this may result in protracted outages and disruption. 

Substantially all operations on an affected engine may be affected:

• Administration of the system using the UI, CLI, or APIs.

• Running or controlling virtual databases (VDBs) or sources).

Contributing Factors

This article applies to the following versions of the Continuous Data Engine:

The issue only affects Continuous Data Engine. 

The issue can only occur during a storage migration job, as started from the CLI using "storage migrate".

Symptoms

An affected engine may become unresponsive:

• The Delphix Admin application will not respond.

• Virtual Databases (VDBs) may become unresponsive and crash.

Relief/Workaround

Customers are advised to defer any block-to-object storage migration operations until a resolution is available for the issue. 

Rebooting an affected system may temporarily restore access; however, the issue is likely to recur.

Related Documents

• In-place block-to-object storage migration