Currently, DV runs with single JVM for all server threads, user requests, caching,etc. and causing too high HEAP requirement, custom GC settings and unrecoverable exceptions. This is causing frequent restarts (due to node shunned alerts) and sometimes metadata corruption.
Since scheduled caching jobs are taking majority of server resources, I'd like to see seperate JVM for each cache [or] cache policy with adjustable JVM configs for heap, GC, etc.
Hi @Guest - rather than introducing separate JVMs for each cache/cache policy, we propose enabling better flexibility over how the systems are allocated to tasks (for example, ensuring that caching tasks don't consume as much memory as user requests). Does this seem like a reasonable solution instead?