August 16, 2022
This release (2.1.0-h1-b10) of Cloudera DataFlow (CDF) on CDP Public Cloud introduces improvements for Prometheus performance, reliability and performance improvements for upgrades, as well as a number of additional bug fixes.
Improvements and fixed issues
- Fixed regression in DFX 2.1.0 where new deployments contain NiFi pods with misconfigured cpu/memory limits. These limits are lower than necessary, leading to the possibility of NiFi pod crashing due to off-heap storage growing large enough to cause an Out of Memory error.
- CDPDFX-5483 - Fixed a bug where disabling DataFlow from certain failure states can potentially leave behind some cloud resources deployed by DataFlow.
- CDPDFX-5486 - Several settings for Prometheus instances running on the workload cluster have been adjusted to improve performance. This includes a reduction in the amount of metrics collected by Prometheus instances and enabling write ahead log compression.
- CDPDFX-5487 - Improved Enable, Upgrade/rollback and Create deployment performance to more reliably interpret status of helm releases and NiFi resources for improved success rates.
- CDPDFX-5494 - Fixed an issue where cluster node scaling events are not being properly surfaced to the user.
- CDPDFX-5495 - Fixed a bug where upgrades failed if a deployment was created with autoscaling disabled, but since was updated in any way (node count, params, and so on).
- CDPDFX-5498 - Improved upgrade timeout configurations to account for delays upgrading clusters with high node counts.
- CDPDFX-5568 - Fixed a bug causing some environments not to show in the filter options on the Deployment Dashboard.
- CDPDFX-5571 - Fixed potential for up to 10-minute delay during upgrades where helm repository sync is unnecessarily delayed. This could sometimes lead to timeout failures during upgrade.
- CDPDFX-5572 - Fixed a bug in the user interface on the Deployment Dashboard when an environment filter option is selected, but becomes no longer valid.
- CDPDFX-5606 - Fixed a bug in retrying a failed DataFlow enablement where a new enable attempt fails fast and old infrastructure is eventually deleted. Old infrastructure details are not cleared from the database, so subsequent retries attempt and fail to clean up infrastructure that is already deleted.
- CDPDFX-5676 - Fixed an issue where in-cluster certificates are renewed, but certain cluster resources are not automatically restarted to start using the new certificates.
- CDPDFX-5736 - Fixed an issue limiting upload of flow definition files to the catalog to 1MB in size. Flow uploads are now supported up to 25MB in size.
- CDPDFX-5780 - Fixed an issue preventing creation of deployments with large flow definitions.