[Openstack-operators] Problems (simple ones) at scale... Invisible VMs.
openstack at medberry.net
Wed May 18 22:13:25 UTC 2016
So, we just ran into an "at-sale" issue that shouldn't have been an issue.
Many of the OpenStack CLI tools accept a limit parameter (to limit how much
data you get back from a single query). However, much less well documented
is that there is an inherent limit that you will run into at a 1000 VMs
(not counting deleted ones). Many operators have already exceeded that
limit and likely run into this. With nova cli and openstack client, you can
simply pass in a limit of -1 to get around this (and though it will still
make paged queries, you won't have "invisible" VMs which is what I've begun
to call the ones that don't make it into the first/default page.
I can't really call this a bug for Nova (but it is definitely a bug for
Cinder which doesn't have a functional get me all of them command and is
also limited at 1000 for a single call but you can never get the rest at
least in our Liberty environment.)
box:~# nova list |tail -n +4 |head -n -1 |wc
1000 16326 416000
box:~# nova list --limit -1 |tail -n +4 |head -n -1 |wc
1060 17274 440960
(so I recently went over the limit of 1000)
-------------- next part --------------
An HTML attachment was scrubbed...
More information about the OpenStack-operators