Uncaught GreenletExit in ProcessLauncher if wait called after greenlet kill

[Carl Baldwin]

I'm running wait in ProcessLauncher in a green thread. I attempted to kill the green thread and then call wait so that the process launcher object can reap its child processes cleanly. This resulted in a trace resulting from a GreenletExit exception being thrown.

The eventlet documentation states multiple times that GreenletExit is thrown after .kill() has been called to kill a thread. I think ProcessLauncher should expect this and deal with it.

[Carl Baldwin (carl-baldwin)]

http://paste.openstack.org/show/67359/

[OpenStack Infra (hudson-openstack)]

Fix proposed to branch: master
Review: https://review.openstack.org/74831

[OpenStack Infra (hudson-openstack)]

Reviewed: https://review.openstack.org/76380
Committed: https://git.openstack.org/cgit/openstack/oslo-incubator/commit/?id=a6fec438abe61d40fc4fb6907ea1fefd4dd93e22
Submitter: Jenkins
Branch: master

commit a6fec438abe61d40fc4fb6907ea1fefd4dd93e22
Author: Carl Baldwin <email address hidden>
Date: Tue Feb 25 22:28:10 2014 +0000

    Reap child processes gracefully if greenlet thread gets killed

    If ProcessLauncher is run from within a thread we should be able to
    call wait after the thread is killed to reap the child processes.
    Catching this exception and then proceeding to the cleanup phase
    accomplishes this.

    Closes-Bug: #1282206
    Change-Id: I9c863864c25c0478b720fc57cf0dccbe9bc9d9f6

[Matt Riedemann (mriedem)]

This was never synced to any projects in Icehouse, the Neutron patch was abandoned.

[Matt Riedemann (mriedem)]

Nova cherry pick to stable/icehouse here: https://review.openstack.org/#/c/94271/

[Alan Pevec (apevec)]

Was this synced to trunk in Juno?

[Matt Riedemann (mriedem)]

@Alan, yeah it's in nova on master for Juno via this change: https://review.openstack.org/#/c/77267/

[OpenStack Infra (hudson-openstack)]

Reviewed: https://review.openstack.org/94271
Committed: https://git.openstack.org/cgit/openstack/nova/commit/?id=705ad64518c76629b9bc7bfb42cb2974667bdf56
Submitter: Jenkins
Branch: stable/icehouse

commit 705ad64518c76629b9bc7bfb42cb2974667bdf56
Author: Carl Baldwin <email address hidden>
Date: Tue Feb 25 22:28:10 2014 +0000

    Reap child processes gracefully if greenlet thread gets killed

    If ProcessLauncher is run from within a thread we should be able to
    call wait after the thread is killed to reap the child processes.
    Catching this exception and then proceeding to the cleanup phase
    accomplishes this.

    Conflicts:
     nova/openstack/common/service.py

    Closes-Bug: #1282206
    Change-Id: I9c863864c25c0478b720fc57cf0dccbe9bc9d9f6
    (cherry picked from commit a6fec438abe61d40fc4fb6907ea1fefd4dd93e22)

[Rossella Sblendido (rossella-o)]

Fix for Neutron was released https://review.openstack.org/#/c/98259/