Closed sharonx closed 6 years ago
Yes idle cpu/memory after containers exit went down as compared to the data before. Starting from 2:30am in the chart ran the test with this fix. Ran it with forced GC, memory went down even further (~2.5% memory usage)
So the memory went down by ~50% ?
@tonytamsf memory went down by ~50% in the graph but could go down even more with forced GC (OS doesn't GC everything until needed).
The root cause of the cpu and memory issue is that we didn't return from the go routines once a container exits (each container launches 2 routines). The buffer used in the go routines can't be garbage collected either.
@sharonx Changes look good.
If extended testing results show reduced cpu/mem profiles - merge.