After having installed FaSS following this guidelines, you need to set the initial shares and the ttl for every user.
$ cd /tmp/one-fass/etc
and edit the file shares.conf
.
Check your OpenNebula users:
oneuser list
and the groups
onegroup list
By default, after the installation of Opennebula, you will have two administrative accounts, oneadmin
(userID=0) and serveradmin
(userID=1). And two groups, oneadmin
(groupID=0) and users
(groupID=1).
The shares.conf
file, therefore, contains the basic working setup for only the administrative accounts and an example of how to add a user called goofy
[oneadmin]
uid=0
gid=0
share=50
ttl=600
[serveradmin]
uid=1
gid=0
share=0
[goofy]
uid=2
gid=1
share=50
ttl=600
you need to change the users accordingly to your list.
To start FaSS, simply type
systemctl start fass
Starting from FaSS v1.1, the scheduler implements a simple version of the SLURM MultiFactor algorithm, setting the parameter plugin_debug
to 1 in the FaSS configuration file.
In order to implement your own fair-share algorithm, edit the BasicPlugin
class, in /tmp/one-fass/src/pm
.
A dummy algorithm, which inverts the priorities recevied from OpenNebula, is also available setting plugin_debug
to 0.
By default the VMs are defined as static, i.e. they will not be terminated.
To have them terminated after they have exceeded their time to live ttl
or maximum waiting time max_wait
, both set in the FaSS configuration file /one-fass/etc/fassd.conf
, you need to add a raw attribute to the onevm instantiate command:
$ onetemplate instantiate <yourtemplateid> --raw static_vm=0
The defalt value for static_vm
is 1
, i.e. the VM is static.
By default the VMs are terminated after the TTL set in the shares configuration file /one-fass/etc/shares.conf
. Other actions are possible, changing the action=kill
in fassd.conf
to be:
poweroff
suspend
reboot
All these actions are performed for VMs in running status.