invoking nova/utils.py's "execute" with check_exit_code=False causes problem in nova-network

[Samuel XJ]

version: essex

In nova/network/linux_net.py, there are six invocations of nova/utils.py's "execute" function with "check_exit_code=False." Our experiences show that this pattern is problematic. Specifically, the process executing a given command may exit unexpectedly.

All examples below are related to nova/network/linux_net.py.

Example 1:
In function "ensure_bridge" of class "LinuxBridgeInterfaceDriver," a command "brctl addif" is executed without checking the exit code. The parent process calling "_execute" cannot know whether the execution of "brctl addif" succeeds. This can cause a bridge and a NIC to remain dissociated.

Example 2:
In function "_device_exists," if the execution of "ip link show dev [device]" fails unexpectedly, then neither stdout nor stderr contains any data. According to the logic in "_device_exists," such a fault would lead the caller of "_device_exists" to think that the device in question exists, which may not be the case.

Example 3:
In function "restart_dhcp," if the execution of "cat /proc/[pid]/cmdline" fails unexpectedly, then there may be no data in the stdout, which in turn will cause the caller to believe that a dnsmasq instance has crashed and a new instance should be started, while in fact the existing dnsmasq instance works properly.

The rest occurrences of "check_exit_code=False" raises similar issues.

[Russell Bryant (russellb)]

This report says Essex, but I just checked the latest code (Grizzly) and the same comments apply there.

[Michael Still (mikal)]

Is this actually true? utils.execute defaults to checking for a zero exit code, and will raise exception.ProcessExecutionError() if the exit code is anything other than that. Its up to the caller to catch the exception.

[Samuel XJ (samuel-xj)]

When the caller of "utils.execute" passes "check_exit_code=False," the
default "[0]" expected exit code list will not be used and the related
exception will not be raised. It is the use of "check_exit_code=False" from
the caller side that we think may be problematic.

On Fri, Mar 15, 2013 at 2:44 PM, Michael Still <email address hidden> wrote:

> Is this actually true? utils.execute defaults to checking for a zero
> exit code, and will raise exception.ProcessExecutionError() if the exit
> code is anything other than that. Its up to the caller to catch the
> exception.
>
> --
> You received this bug notification because you are subscribed to the bug
> report.
> https://bugs.launchpad.net/bugs/1153751
>
> Title:
> invoking nova/utils.py's "execute" with check_exit_code=False causes
> problem in nova-network
>
> To manage notifications about this bug go to:
> https://bugs.launchpad.net/nova/+bug/1153751/+subscriptions
>

[Joe Gordon (jogo)]

This bug is very old and the code has changed significantly since this was filed. If this is still valid in Icehouse or later please reopen.