Skip to content

Latest commit

 

History

History
59 lines (49 loc) · 3.96 KB

CHANGELOG.md

File metadata and controls

59 lines (49 loc) · 3.96 KB
Raw

Changelog

2.0.3

  • Add helm value ingress.uriHttpsProtoOverride to set ingress endpoints with “https” protocol backed by TLS upstream, as opposed to the default TLS configuration using ingress.tls helm values - #7134
  • Fix use of passed in helm value: “pachd.oauthRedirectURI“ - #7133
  • Require populated PG Passwords on Helm Upgrades - #7131
  • Improve large commit finishing performance with autoscaling - #7121
  • Add helm value pachd.activateAuth, to control whether the config-job bootstraps auth on the cluster. Defaults to true - #7136
  • Fixes a potential pachd crash during initialization - #7114
  • Improved the performance of commit finishing and fixed related issues - #7123
  • The sharding and concatenating tasks during compaction are now distributed.
  • The finishing of commits in different repos is now parallelized.
  • Duplicate work that can occur when determining which levels to compact has been removed.
  • File sets created through the file set API are now compacted to improve file set read performance.
  • Fixes a gRPC leak that can occur when collecting info from workers in pachd.
  • Fixes an issue where the job registry would process jobs in the finishing state.
  • Fixes a set of deadlocks that can occur in the postgres listener.

2.0.2

  • Improve tracker GC query performance - #7093
  • Add pachctl delete commit command - #7094
  • Add the “activateEnterprise” helm value that can be set to bootstrap a cluster’s enterprise/auth features during upgrade releases. During installs, providing the enterpriseLicenseKey will trigger the bootstrap process. - #7104
  • Improve logging by disambiguating some commit and job logs - #7111
  • Add helm configuration to register a pachyderm cluster with an external enterprise server. #7109

2.0.1

  • Fixes a goroutine leak that can occur when a modify file operation is canceled - #7055
  • Improves performance by reducing the number of postgres requests needed when renewing chunks being uploaded - #7052
  • Improves usability by changing helm defaults- activate enterprise automatically and use default localhost URIs when ingress is disabled - #7039
  • Improves memory scaling of join and group inputs beyond millions of datums and - - - - Increases logging verbosity by including etcd and pg-bouncer logs in debug dumps - #7053

2.0.0

Introducing Pachyderm 2.0 with several foundational improvements. Read more details here

What’s new

  • New storage architecture and FileSets for better support for small files, content defined chunking for better de-duplication, automatic compression and encryption of chunks, automatic garbage collection, and more
  • Pachyderm Enterprise Management to allow site-wide configuration
  • New enterprise UI -- Pachyderm Console
  • Improved OAuth-based Authentication
  • Simplified and unified lineage tracking with Global IDs
  • More efficient job run time to significantly reduce job completion time for real world workloads
  • Introducing Pachyderm deployment support using Helm Chart
  • Updated python-pachyderm 7.0 release which supports Pachyderm 2.x

Changes from Pachyderm 1.x behavior

  • Empty directories are no longer supported
  • Default upload behavior changes from append to overwrite
  • Full paths in repo must be specified when uploading files
  • Automatic file splitting is no longer supported
  • pachctl deploy has been deprecated in favor of helm charts. See docs for deployment details
  • Standby option replaced by autoscaling
  • Writing multiple input datums to same output file is no longer supported and will result in an error

Updated docs are at docs.pachyderm.com

Try Pachyderm 2.0 on hub.pachyderm.com

Pachyderm 2.x is not backwards compatible with Pachyderm 1.x data formats. If you require assistance or have any questions, please contact [email protected]