Open Stack

Hi, devs,

I noticed that Ceilometer project uses lots of other OpenStack services,
and other thirdparty services APIs, but rare of them set timeout when call
http request, this is not a good behavior because many pollsters run in one
of threads, if one is every slow or stuckd, then others will not be able to
work too.

The worst case is if Ceilometer thread is stuckd, then when outside service
becomes normal, Ceilometer cannot recover itself, cloud operator will need
to restart the service manually, that is bad.

I have reported a bug, see:
https://bugs.launchpad.net/ceilometer/+bug/1388778
and uploaded a patch, see: https://review.openstack.org/#/c/132974/

What's your opinion?

And *I need help for writing test code for this patch* (the main reason why
I write this mail), is there any idea to test timeout case? I don't want
the unit test to be very slow, so a real timeout like sleep is not my
current choice.

Thanks
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.openstack.org/pipermail/openstack-dev/attachments/20141106/ae80ca1f/attachment.html>