I collected data on an example program from four angles. See https://docs.google.com/document/d/1KUz7IjnD93X2VTVkRAhhNa7rHScCaIH8GLrsoaDIW_g for the raw data and my puny attempts to correlate the four views. There are some exact equalities, some near matches, some gross differences in things that seem like they should be the same, and things I wasn't able to match in any way. This begs lots of questions.
Why is memory.usage_in_bytes from cgroups so much smaller than the other top-line measurements? What is statm.size counting that the coredump "all" is not? I assume that the text part of the connection-agent binary is shared. Why is the readonly data part of the connection-agent binary _not_ shared? What is statm.data counting that MemStats.Sys is not? Why is MemStats.HeapIdle so much bigger than the coredump's heap free spans? Why is MemStats.heapReleased so much bigger than the coredump's heap released? In MemStats, (HeapIdle - HeapReleased) = 1,490,944 In coredump, (heap free spans) - (heap released) = 1,613,824 --- not so very different. What is making the roughly 52 MB difference between MemStats and the coredump? Why is the bss in the coredump so huge? Does it correspond to anything in any of the other views? What do the three big anonymous blocks in the procfs view correpond to in the other views? Thanks, Mike -- You received this message because you are subscribed to the Google Groups "golang-nuts" group. To unsubscribe from this group and stop receiving emails from it, send an email to golang-nuts+unsubscr...@googlegroups.com. To view this discussion on the web visit https://groups.google.com/d/msgid/golang-nuts/7c44fbd8-976a-4bfa-acf4-fa4a0d6bad35o%40googlegroups.com.
