Ranking Datastores


(Omar Othman) #1

I asked on IRC about the complete list of variables to be used for ranking hosts accepting new VMs. This is what I know after a helpful gent answered:
(1) Predefined: http://docs.opennebula.org/5.2/operation/references/template.html#predefined-host-attributes
(2) Custom: https://docs.opennebula.org/5.2/integration/infrastructure_integration/devel-im.html

I lost my second IRC session, so decided it might be better to ask here.

Now my original question was:
(1) What is this list (the variables I can use) for datastores? What can I use for “SCHED_DS_RANK”?
(2) How can I expand this list?

My own answer to (2) – which I want to verify – is probes. Basically every time you execute the probes for a host, that host can give you any number of variables back, and I know there is a requirement that any host in a given cluster must has access to any image or system datastore in that cluster. But if that is the case, are we getting N(umber of hosts) reports for every datastore we have? Or can certain probes be executed only by certain hosts? But what about HA in that case? Does OpenNebula know how to “execute this probe on one host only, I don’t really care which host this is”?


Versions of the related components and OS (frontend, hypervisors, VMs):
OpenNebula 5.2.1, KVM

Steps to reproduce:

Current results:

Expected results: