Description of problem: With latest packages, ceilometer doesn't report metrics [root@localhost ~]# . keystonerc_admin [root@localhost ~(keystone_admin)]# ceilometer sample-list -m memory +-------------+------+------+--------+------+-----------+ | Resource ID | Name | Type | Volume | Unit | Timestamp | +-------------+------+------+--------+------+-----------+ +-------------+------+------+--------+------+-----------+
This appears to be due to the use of independent virt inspectors in Icehouse and in particular the lack of functionality within the libvirt inspector : /var/log/ceilometer/compute.log ~~~ 2014-07-01 10:08:35.630 11293 INFO ceilometer.agent [-] Polling pollster memory.usage 2014-07-01 10:08:35.631 11293 DEBUG ceilometer.compute.pollsters.memory [-] Checking memory usage for instance 2828fdaa-2e71-4e10-9d9b-00694bb2ef01 get_samples /usr/lib/python2.7/site-packages/ceilometer/compute/pollsters/memory.py:31 2014-07-01 10:08:35.632 11293 DEBUG ceilometer.compute.pollsters.memory [-] Obtaining Memory Usage is not implemented for LibvirtInspector get_samples /usr/lib/python2.7/site-packages/ceilometer/compute/pollsters/memory.py:51 ~~~ Eoghan, should we close this as a duplicate of RFE BZ#1042332?
Eoghan, any update on Lee's question? Thanks
I think there's been misunderstanding of the log message referenced by Lee above in Comment 2. Even though the originating python module is called memory.py, the pollster defined therein is responsible for emitting a separate memory.usage meter (as opposed to 'memory'). This 'memory.usage' meter is not supported for the libvirt inspector, hence the expected exception. The 'memory.usage' meter was added specifically for the VMware vSphere inspector introduced in Icehouse. The 'memory' meter, on the other hand, is gathered from notifications emitted by nova (as opposed to being polled from the hypervisor). I am seeing this meter being produced, details below[1]. Can you check whether you've got an ACTIVE instance running and also whether notifications are enabled for nova? [1] $ sudo rpm -qa | grep openstack-ceilometer-compute openstack-ceilometer-compute-2014.1.1-1.el6ost.noarch $ nova list +--------------------------------------+------+--------+------------+-------------+---------------------+ | ID | Name | Status | Task State | Power State | Networks | +--------------------------------------+------+--------+------------+-------------+---------------------+ | c420887c-e8b3-4c65-b30f-a0b47696ba09 | s1 | ACTIVE | - | Running | public=172.24.4.227 | +--------------------------------------+------+--------+------------+-------------+---------------------+ $ ceilometer statistics -m memory +--------+----------------------------+----------------------------+-------+-------+-------+--------+-------+----------+----------------------------+----------------------------+ | Period | Period Start | Period End | Max | Min | Avg | Sum | Count | Duration | Duration Start | Duration End | +--------+----------------------------+----------------------------+-------+-------+-------+--------+-------+----------+----------------------------+----------------------------+ | 0 | 2014-08-01T08:29:36.262000 | 2014-08-01T08:29:36.262000 | 512.0 | 512.0 | 512.0 | 1024.0 | 2 | 240.181 | 2014-08-01T08:29:36.262000 | 2014-08-01T08:33:36.443000 | +--------+----------------------------+----------------------------+-------+-------+-------+--------+-------+----------+----------------------------+----------------------------+
On the specific question of being duplicate of RFE BZ#1042332 ... I wouldn't agree, as the current bug reports the 'memory' meter not being seen, whereas BZ#1042332 is a request for the addition of support to libvirt for a memory utilization meter (along the lines of the VMware vSphere 'memory.usage' meter mentioned in the previous comment).
Do we have ETA for this? Regards, Pablo
Hi Pablo, I think we're still waiting on an answer to the question in Eoghan's first comment: "Can you check whether you've got an ACTIVE instance running and also whether notifications are enabled for nova?" His comment suggests that in order for there to be a memory meter there needs to be at least one active instance and notifications need to be turned on.
Chris, Checking with customer, setting needinfo on me for updating afterwards. Thanks! Pablo
Pablo, any status on this, it's been open and pending needinfo from you for a long time. Can will resolve it?
Chris, customer not updated the case anymore. Thanks for closing this.