shutdown behaviour of rkt containers

gistfile1.txt

x

reaper.sh

#!/usr/bin/bash
shopt -s nullglob

SYSCTL=/usr/bin/systemctl

# just adding some debug to the reaper
echo "in the reaper"

if [ $# -eq 1 ]; then
    app=$1
    ${SYSCTL} status "${app}.service" >> /rkt/reaper.log

    status=$(${SYSCTL} show --property ExecMainStatus "${app}.service")

    echo "${status#*=}" > "/rkt/status/$app"
    if [ "${status#*=}" != 0 ] ; then
        # The command "systemctl exit $status" sets the return value that will
        # be used when the pod exits (via shutdown.service).
        # This command is available since systemd v227. On older versions, the
        # command will fail and rkt will just exit with return code 0.
        ${SYSCTL} exit ${status#*=} 2>/dev/null
    fi
    exit 0

fi

rkt run.txt

if i run a tester container, and `kill -INT 5` (the `ugh` process), then the container exits properly and the reaper is called

```
root@vagrant-ubuntu-trusty-64:~# rkt run sha512-55af61e90580fb573666d07b3a9c6d09 --interactive
image: using image from file /usr/local/bin/stage1-coreos.aci
image: using image from the store with hash sha512-55af61e90580fb573666d07b3a9c6d098dabc94816681c14495f130ada598a85
networking: loading networks from /etc/rkt/net.d
networking: overriding "default" network with 20-default.conf
networking: loading network default with type bridge
starting up
got a signal! interrupt
bye bye
in the reaper
root@vagrant-ubuntu-trusty-64:~#
```

second rkt run

if i run the tester container, and `kill -INT 1` (the init process), then the container exits poorly and the reaper isn't called
```
root@vagrant-ubuntu-trusty-64:~# rkt run sha512-55af61e90580fb573666d07b3a9c6d09 --interactive
image: using image from file /usr/local/bin/stage1-coreos.aci
image: using image from the store with hash sha512-55af61e90580fb573666d07b3a9c6d098dabc94816681c14495f130ada598a85
networking: loading networks from /etc/rkt/net.d
networking: overriding "default" network with 20-default.conf
networking: loading network default with type bridge
starting up
got a signal! hangup
Failed to retrieve machine ID: No such file or directory
```
the interesting thing about that 'machine-id' error, is that the machine-id actually exists (stage1/rootfs/etc/machine-id) while the pod is running, but seems to be removed at some point in the shutdown process

it's also possible that the machine-id error doesn't manifest, for example, this run
```
root@vagrant-ubuntu-trusty-64:~# rkt run sha512-55af61e90580fb573666d07b3a9c6d09 --interactive
image: using image from file /usr/local/bin/stage1-coreos.aci
image: using image from the store with hash sha512-55af61e90580fb573666d07b3a9c6d098dabc94816681c14495f130ada598a85
networking: loading networks from /etc/rkt/net.d
networking: overriding "default" network with 20-default.conf
networking: loading network default with type bridge
starting up
in the reaper
got a signal! hangup
root@vagrant-ubuntu-trusty-64:~#
```

tester.manifest

{
    "acKind": "ImageManifest",
    "acVersion": "0.7.4",
    "app": {
        "exec": [
            "/usr/bin/ugh"
        ],
        "group": "0",
        "user": "0"
    },
    "dependencies": [
        {
            "imageName": "registry-1.docker.io/library/ubuntu",
            "labels": [
                {
                    "name": "os",
                    "value": "linux"
                },
                {
                    "name": "arch",
                    "value": "amd64"
                },
                {
                    "name": "version",
                    "value": "14.04.2"
                }
            ]
        }
    ],
    "labels": [
        {
            "name": "arch",
            "value": "amd64"
        },
        {
            "name": "os",
            "value": "linux"
        }
    ],
    "name": "tester"
}

ugh.go

package main

import (
	"fmt"
	"os"
	"os/signal"
	"syscall"
	"time"
)

func main() {
	handler := make(chan os.Signal, 1)
	signal.Notify(handler, syscall.SIGPIPE, syscall.SIGCHLD, syscall.SIGHUP, syscall.SIGALRM, syscall.SIGINT, syscall.SIGTERM)

	go func() {
		for sig := range handler {
			fmt.Printf("got a signal! %v\n", sig)
			time.Sleep(20 * time.Second)
			fmt.Println("bye bye")
			os.Exit(1)
		}
	}()

	fmt.Println("starting up")
	time.Sleep(1 * time.Hour)
}