"Usage" Metrics

[andreidum]

Hi,
Can you please give more details regarding how the "Used CPU" and "Used memory" metrics are collected and what do they actually mean? For instance, in your own screenshots, there are 8GHz allocated, but only 433M used. In my setup, I have 9 Ghz allocated and 622 Mhz used (cur/avg/max).
Similarly, I have 16 GB of RAm allocated, but only 8.05/8.05/8.13 MB (cur/avg/max) used.
This seems quite odd.
I also stress tested a VM (CPU & memory) and these numbers didn't change a bit.

Thanks

[cluther]

The "used" values are aggregated from calls to the listsHosts API method. So for the entire cloud it should reflect the total usage from all hosts. For a zone it should reflect the usage from only hosts in that zone, and so on.

http://download.cloud.com/releases/2.2.0/api_2.2.12/global_admin/listHosts.html

The 433MHz used CPU sounds somewhat plausible without knowing anything other than that you stress tested one VM. However, the memory used does seem suspiciously low. How many hosts do you have? Would it be possible for you to go through each of those and see what they report for used CPU and memory and see if it makes sense?

[andreidum]

I have 4 hosts and 5 VMs.

If I check the Memory graph of one host, I see 2.36M/4.38G/6.47G (Used/Allocated/Total)
If I ssh to the host, free -m is reporting:

         total       used       free     shared    buffers     cached

Mem: 6723 6607 115 0 105 4084
-/+ buffers/cache: 2417 4306

On a VM running on this host, I get:
free -m
total used free shared buffers cached
Mem: 3016 309 2707 0 37 149
-/+ buffers/cache: 121 2894

[cluther]

Sorry. I meant to check the host CPU and memory usage graphs within Zenoss.

[andreidum]

So, in the Zenoss->Components->Clusters->Cluster1 graphs/view, I see:
Memory: Total 21.4G, Allocated 16.38G, Used: 8.33M/8.06M/8.68M (cur/avg/max)

Since I have a single cluster with 1 zone and 1 pod, the graphs in the pod & zone view are identical:
Total 21.4G, Allocated 16.38G, Used: 8.33M/8.06M/8.68M (cur/avg/max)

In the Components->Hosts view:
cloud01: Total 6.57G, Allocated 4.38G, Used: 2.35M/2.36M/2.43M
cloud02: Total 3.61G, Allocated 3.00G, Used: 1.03M/1.02M/1.03M
cloud03: Total 7.62G, Allocated 6.00G, Used: 2.95M/2.98M/3.19M
cloud04: Total 3.61G, Allocated 3.00G, Used: 2.04M/1.71M/2.06M

I do not collect detailed ZenOSS graphs (eg. via SNMP) for hosts and VMs.

[cluther]

At least it's consistent. The code that's collecting this is pretty straightforward. Unless the memoryused and memorytotal responses to the listHosts API call are represented in different units, I don't see where the problem could be.

I'm seeing if anyone in #cloudstack has any ideas.