CMS: concurrent mode failure - not parallel?
I am currently trying to avoid CMS mode failures in our application. However when they occur (mispredictin or fragmentation) I want the full collection happen as quickly as possible. On the Hotspot VMs I have seen I only see a single thread doing the collection (which takes in my case up to 10s due to large heap). Is there a way to use all 8 cores for that collection type?