from linux kernel/libc chronicle java runtime.exec() stable regards memory?


at work the aim platforms apparatus compelled mini-server controlling linux (kernel 2.6.13, tradition chain formed an aged fedora core). concentration combined java (sun jdk 1.6_04). linux oom torpedo configured kill processes memory use exceeds 160mb. even during high bucket the concentration never over 120mb together internal processes active stay good within oom limit.



however, turns out java runtime.getruntime().exec() method, certified proceed govern outmost processes java, causes spawned child processes (temporarily) need same volume memory progenitor slight given castle space copied. net outcome the concentration gets killed oom torpedo shortly runtime.getruntime().exec().



we now work around carrying removed internal way outmost management execution promulgate way over socket. reduction optimal.



after i got feedback indicating should start "newer" versions linux given exercise posix fork() slight controlling copy-on-write, presumably definition wholly duplicate pages needs cgange mandatory instead whole castle space immediately.



my questions are:




  • is true?

  • is something kernel, libc doing somewhere else entirely?

  • from chronicle kernel/libc/whatever copy-on-write fork() available?



Comments

Popular posts from this blog

list macos calm editors formula editors

how i practical urls indicate .aspx pages asp.net deployed an iis? (preferably but iis)

jaxb - xjc - reworking generated typesafe enum category members