Given that, I'm marking the kernel tasks as invalid as this appears to be stress-ng being over aggressive.
MAAS deployments (and I presume server installs in general) only provide about 4 - 8GB swap max in a swap file created in the root filesystem (no swap partition anymore). It is not reasonable to create, IMO, 100GB swap files just to be sure we don't accidentally trip the OOMKiller when running memory tests. Also, the ratio seems to be variable. On my amd64 system, the swap consumed was 5x the amount of installed RAM. On the ppc64le system, the amount was just shy of 2x the amount of physical RAM (this is also a NUMA capable system, so maybe that's a contributing factor, beyond the differences in how each arch handles memory, which I freely admit to not understanding well. -- You received this bug notification because you are a member of Ubuntu Bugs, which is subscribed to Ubuntu. https://bugs.launchpad.net/bugs/1729878 Title: stress-ng triggering oomkiller running brk and stack stressors on all arches. To manage notifications about this bug go to: https://bugs.launchpad.net/stress-ng/+bug/1729878/+subscriptions -- ubuntu-bugs mailing list [email protected] https://lists.ubuntu.com/mailman/listinfo/ubuntu-bugs
