openstack team mailing list archive

[이준원 (Joonwon Lee) <joonwon7.lee@xxxxxxxxxxx>]

Thanks for swift answer, Vish. :)

I can completely agree that using external monitoring services is better for the OpenStack. I thought that it would be nicer for Nova to provide the same monitoring APIs for libvirt and xenapi (and others). But, now I understand it's not so easy task to accomplish.
Maybe only we are preparing monitoring facilities in the middle of considering and choosing between KVM and XenServer. :)

Regards

Joonwon Lee



From: Vishvananda Ishaya [mailto:vishvananda@xxxxxxxxx] 
Sent: Thursday, September 29, 2011 7:24 PM
To: 이준원 (Joonwon Lee)
Cc: openstack@xxxxxxxxxxxxxxxxxxx
Subject: Re: [Openstack] Why did the instancemonitor disappear in the Diablo?

They were removed because they no longer worked correctly and they were based on twisted code.  There is no current plan to replace them.  Rather than generating graphs via internal software, it makes more sense to write very simple plugins for external monitoring services like munin.

Example using nova db calls:

https://github.com/cloudbuilders/openstack-munin/blob/master/nova_floating_ips

Example using custom sql:

https://github.com/cloudbuilders/openstack-munin/blob/master/nova_instance_timing

Existing ones for network and block monitoring:

http://honk.sigxcpu.org/projects/libvirt/

Vish

On Sep 29, 2011, at 2:58 AM, 이준원 (Joonwon Lee) wrote:


Dear All

Why did the instancemonitor disappear in the Diablo release?
For example, following files can be found in the Diablo milestone-3. But
they're removed in the Diablo milestone-4 and the last release.
   bin/nova-instancemonitor
   nova/compute/monitor.py

Can someone explain the reason why these are removed and whether there is
any plan to replace them.

Joonwon Lee