Who can we report this to ? That other guy who opened the bug didn't get anywhere and this seems pretty serious too. It's the second major but to affect ZFS this year and frankly it makes FreeBSD seem unsuitable for production..
Who can we report this to ? That other guy who opened the bug didn't get anywhere and this seems pretty serious too. It's the second major but to affect ZFS this year and frankly it makes FreeBSD seem unsuitable for production..
i used a malloc bomb to try to mitigate such a problem in the past. you need to sysctl vm.overcommit to 1 and maybe disable swap if it is on spinning rust. it sometimes worked but the machine had only 1gb ram and in the end i moved over
the machine (was a vm in fact) was the target of zfs send and sometimes zfs recv failed with out of memory. the purpose of the bomb was to force arc/wired to a lower level.
it had no swap and it only ran sshd.
the bomb was something like
#include <stdlib.h>
#include <unistd.h>
int main() {
while(malloc(1024*1024*64)) usleep(100);
return 0;
}
the sleep was so i can see something on vm monitoring tools but most likeley is not needed
This site uses cookies to help personalise content, tailor your experience and to keep you logged in if you register.
By continuing to use this site, you are consenting to our use of cookies.