On 11/11/15 at 07:48pm, Jeremy Stanley wrote:
On 2015-11-11 11:25:09 -0600 (-0600), Chris Friesen wrote:
I didn't think that the overhead of deleting/creating an instance was *that*
much different than rebuilding an instance.
Do you have any information about where the "significant performance
advantage" was coming from?
The main reason I recall the suggestion coming up is that, due to
IPv4 address starvation in some provider regions, nova boot was
taking an hour or more to return waiting for an IP address to be
assigned. Using nova rebuild would have supposedly avoided the
address churn and thus improved instance turnaround for us.
There may also have been other reasons for recommending rebuild to
us, but if so I don't recall what they were.
In addition to waiting for IP allocation the significant differences
between boot and rebuild are likely to be around scheduling and
potential build failures. Scheduling should take seconds at most and
shouldn't be a significant factor in the general case. However,
differences between the scheduler view of resources and the compute view
of resources means there's a greater chance of a build failing on a
compute host, and needing a reschedule and potentially another image
download to a host, than a rebuild.
Under ideal conditions there should not be much of a difference between
a fresh boot and a rebuild, but booting a fresh instance has more edge
cases that can be hit.
--
Jeremy Stanley
__________________________________________________________________________
OpenStack Development Mailing List (not for usage questions)
Unsubscribe: openstack-dev-requ...@lists.openstack.org?subject:unsubscribe
http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-dev
__________________________________________________________________________
OpenStack Development Mailing List (not for usage questions)
Unsubscribe: openstack-dev-requ...@lists.openstack.org?subject:unsubscribe
http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-dev
