With the third edition of CPU-Pooler it becomes possible to finally, physically separate the CPUs of all your workloads, while also making it possible to run CPU-Pooler on all your nodes!
The following Changelog describes the fixes and usability improvements released with v0.3.0:
- Project now supports a third CPU pool type, default. Workloads not defining explicit CPU pools are automatically connected to default:
#8 - CPUSetter is now working properly, and automatically isolates the cpuset of all your workloads based on their CPU pool requests!
#9 - CPUSetter now properly reset a container's cpuset if it was restarted for any reason:
#9 - It is now possible to configure the CPU pools with the standard Linux notation, including discontinuous ranges:
#12 - Invalid CPU pools are not advertised anymore to the Device Manager:
#10 - Cpu.requests is now zeroed out for shared pool users to avoid double bookkeeping of the same CPU resource:
#14
After these changes CPU-Pooler is now officially ready for production!
Like the project? Try it out and get back to us with your feedback!