cannot allocate memory errors

Hi,

We have an instance where a controller has free memory, but juju keeps logging this kind of errors :

2017-03-08 14:13:37 ERROR juju.worker.dependency engine.go:547 "disk-manager" manifold worker returned unexpected error: cannot list block devices: lsblk failed: fork/exec /bin/lsblk: cannot allocate memory

$ free -m
              total used free shared buff/cache available
Mem: 7982 6934 218 83 829 669
Swap: 0 0 0

I think that's because juju will use exec.Command() to run external commands, and this uses a simple clone() as show below from an strace -e process :

[pid 13429] clone(child_stack=0, flags=SIGCHLD) = -1 ENOMEM (Cannot allocate memory)

Using clone() this way means that the address space of the calling process (jujud) will be copied, and jujud is using a hefty amount of memory :

USER PID %CPU %MEM VSZ RSS TTY STAT START TIME COMMAND
root 13429 360 28.2 4031016 2311676 ? Sl 09:48 974:17 /var/lib/juju/tools/machine-1/jujud machine --data-dir /var/lib/juju --machine-id 1 --debug

This was already reported and discussed in https://bugs.launchpad.net/juju-core/+bug/1382556, but for juju 1.25, and apparently exclusively for "juju run", which is different from what I'm reporting, since this is 2.1.1 and not caused by a "juju run".

With the new juju2 way of having controllers host multiple models, increasing the resources used by jujud, this bug is, presumably, more likely to appear.

Thank you