Inaccurate totalRAMUsed reported by "nova absolute-limits"

asked 2015-02-10 17:33:56 -0500

mickael gravatar image

Hello,

I am running OpenStack IceHouse on the top of 4 nodes (1 controller, 1 network-neutron-, 2 compute). Everything is working fine since a couple a weeks, however I have just noticed that the amount that is reported for the "RAM used" by the dummy tenant I am using is inaccurate. The problem exist in Horizon as well as when using nova CLI (nova absolute-limits).

To do it short, here is my problem:
- I have terminated ALL my instances --> here is the output from "nova list":

 [root@controller ~(keystone_userA)]# nova list
+----+------+--------+------------+-------------+----------+
| ID | Name | Status | Task State | Power State | Networks |
+----+------+--------+------------+-------------+----------+
+----+------+--------+------------+-------------+----------+
  • and now here is the output from "nova absolute-limits":
[root@controller ~(keystone_userA)]# nova absolute-limits<br>
+-------------------------+-------+
| Name                    | Value |
+-------------------------+-------+
...
| totalRAMUsed            | 384   |
| totalFloatingIpsUsed    | 0     |
| totalInstancesUsed      | 0     |
| totalCoresUsed          | 0     |
...
+-------------------------+-------+

As you can see, no instance is currently running however 384 (MB) of RAM are reported as used.

In order to try to fix this issue, I have gone onto each of the 2 compute nodes (ssh into them), and run "virsh list --all". Surprisingly, I could find 3 "old" instances in state 'shut off' but still defined. I have manually undefined all of them from within virsh ... and finally, rebooted the 2 compute nodes as well as the controller one.
Unfortunately, this did not fixed the issue about the amount of "RAM used" reported wrongly to 384 MB.

Is there anyone who could help in getting the amount of "RAM used" reported back to 0?

Thanks in advance for any help you could provide.

edit retag flag offensive close merge delete