0.6.0 release of Tempo comes with lots of features, enhancements and bugfixes. 🎁 🚀
The highlights of this release include support for Compression and Exhaustive search, along with CPU performance gains from changing compression algorithm between distributors and ingesters.
Breaking changes
[CHANGE] Ingester cut blocks based on size instead of trace count. Replace ingester traces_per_block setting with max_block_bytes. This is a breaking change. #474 (By default, Tempo will cut blocks of size 1GB) @mdisibio
[FEATURE] Added block compression. This is a breaking change b/c some configuration fields moved. #504 (By default, Tempo will use zstd for compression.) @joe-elliott
[CHANGE] Refactor cache section in tempodb. This is a breaking change b/c the cache config section has changed. #485 @dgzlopes
Enhancements
[CHANGE/BUGFIX] Rename tempodb_compaction_objects_written and tempodb_compaction_bytes_written metrics to tempodb_compaction_objects_written_total and tempodb_compaction_bytes_written_total. #524 @gouthamve
[ENHANCEMENT] Change default ingester_client compression from gzip to snappy. #522 @mdisibio
[ENHANCEMENT] Add exhaustive search to combine traces from all blocks in the backend. #489 @annanay25
[ENHANCEMENT] Add per-tenant block retention #77 @mdisibio
[ENHANCEMENT] Change index-downsample to index-downsample-bytes. This is a breaking change #519 @joe-elliott
Bugfixes
[BUGFIX] Upgrade cortex dependency to v1.7.0-rc.0+ to address issue with forgetting ring membership #442 #512 @mdisibio
[BUGFIX] No longer raise the tempodb_blocklist_poll_errors_total metric if a block doesn't have meta or compacted meta. #481 @joe-elliott
[BUGFIX] Replay wal completely before ingesting new spans. #525 @joe-elliott
For the full list of changes see CHANGELOG