[ovirt-users] 答复: 答复: How power-saving schedule policy applied?

2018-01-09 Thread Alex Shen
Martin, 
Thanks for your response.
I've understood why ovirt keep hosts resource reserved. Meanwhile, I totally 
agree with you that CLI is better manner to do timeline based power saving 
policy.

Best regards


> Set ‘HostInReserve’ to 0. Bingo, cloudstack03 goes down as expected.
>
> But when I launch a new VM I get an error popup dialog which inform me 
> that there isn’t enough memory resource to start a new VM. Is it 
> normal way that power saving mechanism works? Can’t power saving policy 
> scheduling module wake up a host to launch VM automatically?
>

We decided against this. Booting a host can take a long time + there is some 
time before our management agent reports in and we can't wait that long when 
starting a VM. That is why we keep the empty host in reserve. That way the VM 
can be started immediately and the additional host's boot time does not matter 
too much.

> Also, Is there any timeline base power saving policy I can apply? Like 
> from 8’clock AM to 6’clock PM, 10 hosts up, 6’clock PM to next 8’clock AM, 
> just keep 3 hosts alive.

We do not have UI for this, but it is easily done using cron and REST API 
(either directly or using some SDK).
We were even working on it for a while, but it was put on backburner since 
sysadmins know cron well and the UI would have to be limited anyway.

Best regards

--
Martin Sivak
SLA / oVirt

On Tue, Jan 9, 2018 at 8:55 AM, Alex Shen <a...@bill-jc.com> wrote:
>
> Hi Martin, Karli
>
>
>
> Thank a lot for your help.
>
>
>
> According to your instruction, I added two parameters into cluster 
> configuration, as below.
>
> 1)  enable power cycling mechanisms
>
> 2)  reserve one host standing by
>
> My test bed:
>
> 3 hosts which are real servers, cloudstack01/cloudstack03/cloudstack04.
>
> Heavy load on cloudstack01, 21 vms
>
> Light load on cloudstack03, 1 vms
>
> No thing on cloudstack04
>
>
>
> When I shut down the last vm on cloudstack03(because I can’t migrate it to 
> cloudstack01, I guess the load is already exceeding the threshold), the 
> cloudstack04 goes to ‘maintenance’ then goes to ‘down’ state. Yepp, POWER 
> SAVING.
>
>
>
>
>
> I tried further test:
>
> Set ‘HostInReserve’ to 0. Bingo, cloudstack03 goes down as expected.
>
> But when I launch a new VM I get an error popup dialog which inform me that 
> there isn’t enough memory resource to start a new VM. Is it normal way that 
> power saving mechanism works? Can’t power saving policy scheduling module 
> wake up a host to launch VM automatically?
>
>
>
> Also, Is there any timeline base power saving policy I can apply? Like from 
> 8’clock AM to 6’clock PM, 10 hosts up, 6’clock PM to next 8’clock AM, just 
> keep 3 hosts alive.
>
>
>
> Best regards
>
> Alex
>
>
>
>
>
> 发件人: Martin Sivak [mailto:msi...@redhat.com]
> 发送时间: 2018年1月8日 19:43
> 收件人: Karli Sjöberg <ka...@inparadise.se>
> 抄送: Alex Shen <a...@bill-jc.com>; users <users@ovirt.org>
> 主题: Re: [ovirt-users] How power-saving schedule policy applied?
>
>
>
> Hi,
>
>
>
> and here is a fresh screenshot just for you :)
>
>
>
> You need to edit the cluster, select scheduling policy tab and add two 
> parameters to your Power saving policy. One enables the power cycling 
> mechanisms and the second one (HostsInReserve) controls how many empty hosts 
> are allowed to stay up. When the host is not empty anymore a new one will be 
> started.
>
>
>
> Best regards
>
>
>
> --
>
> Martin Sivak
>
> SLA / oVirt
>
>
>
> On Mon, Jan 8, 2018 at 12:29 PM, Karli Sjöberg <ka...@inparadise.se> wrote:
>
>
>
>
>
> Den 8 jan. 2018 12:07 skrev Alex Shen <a...@bill-jc.com>:
>
> Hi,
>
>
>
> I’m wondering how to apply power-saving schedule in ovirt cluster(version is 
> 4.2.0.2-1.el7.centos)? Is there any instruction or manual with UI snapshots 
> guidance? That will be appreciated.
>
>
>
> I’ve configed all host with ipmilan protocol and test ok. I suppose 
> that power saving policy should converge VMs into hosts one by one. 
> Something like that. From the beginning, all hosts are in ‘power off’ 
> state, if a VM launch, power saving scheduling module should power on 
> one host, and assign VM into that host. With many VMs activated, 
> exceeding the threshold, power saving scheduling module should turn on 
> another host…
>
>
>
> Jepp, that's basically how it works:)
>
>
>
> /K
>
>
>
>
>
> If anyone had played power saving scheduling policy, please share me your 
> scenarios or some hints. Thanks a lot.
>
>
>
> Alex
>
>
>
> 无病毒。www.avast.com
>
>
>
> ___
> Users mailing list
> Users@ovirt.org
> http://lists.ovirt.org/mailman/listinfo/users
>
>
>
>
> ___
> Users mailing list
> Users@ovirt.org
> http://lists.ovirt.org/mailman/listinfo/users
>
>
>
>
> 无病毒。www.avast.com



---
Avast 防毒软件已对此电子邮件执行病毒检查。
https://www.avast.com/antivirus


___
Users mailing list
Users@ovirt.org
http://lists.ovirt.org/mailman/listinfo/users


[ovirt-users] How power-saving schedule policy applied?

2018-01-08 Thread Alex Shen
Hi,



I'm wondering how to apply power-saving schedule in ovirt cluster(version is
4.2.0.2-1.el7.centos)? Is there any instruction or manual with UI snapshots
guidance? That will be appreciated.



I've configed all host with ipmilan protocol and test ok. I suppose that
power saving policy should converge VMs into hosts one by one. Something
like that. From the beginning, all hosts are in 'power off' state, if a VM
launch, power saving scheduling module should power on one host, and assign
VM into that host. With many VMs activated, exceeding the threshold, power
saving scheduling module should turn on another host.



If anyone had played power saving scheduling policy, please share me your
scenarios or some hints. Thanks a lot.



Alex



---
Avast 防毒软件已对此电子邮件执行病毒检查。
https://www.avast.com/antivirus
___
Users mailing list
Users@ovirt.org
http://lists.ovirt.org/mailman/listinfo/users