Add "cluster" label for hosts, vms and datastores

[pryorda]

From @dannyk81 on February 9, 2018 22:35

It would be great to add "cluster" label to indicate to which HA cluster the host/vm belongs to.

Same, for datastores that belong to a Datastore cluster.

We manage multiple clusters with the same vCenter, and would like to filter metrics per "PROD" and "STAGING" clusters, for example.

Copied from original issue: rverchere/vmware_exporter#28

[pryorda]

From @dannyk81 on February 16, 2018 14:37

@rverchere looking at pyvmomi and vsphere api, I believe there is no simple way to add this label.

But, we should be able to create a new set of metrics with this information, for example:

vmware_vm_dc_cluster{vm_name="vm1",dc_name="dc1",cluster_name="cluster1"} 1
vmware_host_dc_cluster{host_name="host1",dc_name="dc1",cluster_name="cluster2"} 1
vmware_datastore_dc_cluster{ds_name="ds1",dc_name="dc1",cluster_name="cluster3"} 1

Taking some inspiration on how to retrieve this data from here: https://github.com/vmware/pyvmomi-community-samples/blob/master/samples/getvmsbycluster.py

Once we have these metrics in place, we can use the PromQL grouping capability to build queries like so:

(sum(vmware_host_cpu_usage) / sum(vmware_host_cpu_max) * 100) * ON (host_name) group_left(cluster_name) vmware_host_cluster{cluster_name="PROD"}

The above would be very valuebale to us, as it will allow us to aggregate metrics based on the relevant clusters, espeically for alerting since Prod and Staging hosts/vms/datastores are not to be treated the same.

/edit: I think it would also make sense to include the datacenter name as well (examples updated above), also needless to say but these should only be collected from vcenter target(s).

[pryorda]

From @dannyk81 on March 2, 2018 19:31

@rverchere sorry for nagging 😄 just wondering if you had a chance to look into this?

[pryorda]

From @rverchere on March 2, 2018 19:34

Hi,

Sorry but I'm in holidays right now, will have a look on 2 weeks ;)

[pryorda]

From @dannyk81 on March 2, 2018 20:4

Hey! enjoy your time off 😄

[pryorda]

From @jgrumboe on March 7, 2018 10:49

+1 DC and cluster would be really helpful.
Also enjoy your time off.

[pryorda]

From @dannyk81 on April 7, 2018 0:30

ping @rverchere 😄

[pryorda]

From @rverchere on April 7, 2018 8:35

@dannyk81 pong : https://github.com/rverchere/vmware_exporter/blob/master/README.md#unmaintained

Sorry but I started new job with new monitoring tools, so I cannot continue maintain this exporter (yet).

[pryorda]

From @dannyk81 on April 7, 2018 10:57

Best of luck @rverchere!

Hope someone is found to continue this work 🙏

[pryorda]

From @dannyk81 on May 2, 2018 1:48

@rverchere I know you're no longer working on this, but perhaps you will consider to include rverchere/vmware_exporter#36

It provides a meta metric with labels that describe the datacenter, cluster and host per VM.

It can be easily used in PromQL to filter other metrics based on these parameters.

[dannyk81]

Hey @pryorda! I've seem to have missed you taking over, apologies!

I'm going to try and prepare a PR to add this into your variant, may take me some time as I see quite a bit has changed.