[jira] [Updated] (CLOUDSTACK-8181) Overprovisioning hosts happens when using first fit allocator + high disable threshold

["Nitin Mehta (JIRA)" <ji...@apache.org>]

     [ https://issues.apache.org/jira/browse/CLOUDSTACK-8181?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]

Nitin Mehta updated CLOUDSTACK-8181:
------------------------------------
    Description: 
1) In the host allocator, under a given cluster, we list all the hosts which match host tags
2) We iterate over this list of hosts and for each host check that the host has capacity or not
3) We return the first host found having capacity.
4) This causes usually the same host to get all the VMs until it goes out of capacity
5) So the change needed is return the host which has maximum free capacity from the list of hosts, instead of the first fit


  was:
1) In the host allocator, under a given cluster, we list all the hosts which match host tags
2) We iterate over this list of hosts and for each host check that the host has capacity or not
3) We return the first host found having capacity.
4) This causes usually the same host to get all the VMs until it goes out of capacity
5) So the change needed is return the host which has maximum free capacity from the list of hosts, instead of the first fit
6) Since this is not a fix but kind of a mini-feature, we need QA tests to check that the order used by the allocator is correctly working.


> Overprovisioning hosts happens when using first fit allocator + high disable threshold
> --------------------------------------------------------------------------------------
>
>                 Key: CLOUDSTACK-8181
>                 URL: https://issues.apache.org/jira/browse/CLOUDSTACK-8181
>             Project: CloudStack
>          Issue Type: Improvement
>      Security Level: Public(Anyone can view this level - this is the default.) 
>            Reporter: Nitin Mehta
>
> 1) In the host allocator, under a given cluster, we list all the hosts which match host tags
> 2) We iterate over this list of hosts and for each host check that the host has capacity or not
> 3) We return the first host found having capacity.
> 4) This causes usually the same host to get all the VMs until it goes out of capacity
> 5) So the change needed is return the host which has maximum free capacity from the list of hosts, instead of the first fit



--
This message was sent by Atlassian JIRA
(v6.3.4#6332)