-Docs @ v0.11.0
-Examples @ v0.11.0
Changes
Features
- ✨ Add log stored status explicitly (#760)
- Added log stored stored status explictly in the Log object improves the detection for partial or no storage of logs
- ✨ Enable storing of Run Events as Record (#748)
All Events related to taskrun or PipelineRun are stored when we are done with Runs and in a single List.
This can be controlled by a flag passed to the watcher: "store-event". Setting it to false
disables the storing of Events.
The record Name of EventList is stored as results.tekton.dev/eventlist
in TaskRun and PipelineRun.
Fixes
- 🐛 move log storage off reconciler thread (again) (#759)
Prior attempts to fix the watcher memory leak when log storage is enabled (originally issue #695) placed log storage on the dynamic reconciler to facilitate retry on retryable errors. However, production level testing uncovered underlying golang/http2 performance issues which made this untenable.
This change moves log storage back off the dynamic reconciler thread, but still addresses the memory leak, as well as intermittent 'canceled context' errors during log storage through the results api server.
Two new command line arguments are enabled on the watch to facilitate timeouts on log storage:
- dynamic_reconcile_timeout: How long the Watcher waits for the dynamic reconciler to complete before it aborts
- update_log_timeout: How log the Watcher waits for the UpdateLog operation for storing logs to complete before it aborts.
- 🐛 do not lose flush error on server side update log (#756)
Misc
- 🔨 Use latest Kind k8s version, fixes e2e tests (#766)
Thanks
Thanks to these contributors who contributed to v0.11.0!
- ❤️ @enarha
- ❤️ @gabemontero
- ❤️ @khrm
Extra shout-out for awesome release notes:
- 😍 @enarha
- 😍 @gabemontero
- 😍 @khrm