The presentation discusses the journey towards the GA release of CronJob API in Kubernetes, including the challenges faced and the improvements made in the new controller.
- The idea of CronJob API was introduced in Kubernetes in 2015.
- The initial implementation of the scheduled job controller was challenging due to the lack of API groups and versions.
- The old controller had performance and scaling issues when handling thousands of cron jobs.
- The new controller solved these issues and introduced new features such as a histogram metrics and minor optimizations.
- The new controller was tested with a sample workload of 6120 cron jobs and showed no visible delay in scheduling.
- The presentation expresses gratitude towards the users and developers who contributed to the development of the CronJob API.
The old controller had scaling problems when adding more and more cron jobs, which resulted in a linear increase in time between processing consecutive job creation. In contrast, the new controller did not experience any lag in job creation even when tested with a workload of 6120 cron jobs. This illustrates the significant performance boost and improvements made in the new controller.