2

First of all, excuse my non acquaintance in IT, because I'm not an IT person at all. I just installed a small render farm composed of 16 HP proliant 380 G5 and 10 Bull Novascale R2 (supermicro based motherboard). OS is windows server 2008 R2. I would like to switch on or shutdown all the servers from another Win7 machine. The win7 machine is now on the same local network, but I'm looking for a solution that can work even outside the local network. The G5 are provided with Ilo 2 while the bull used supermicro IPMI.

On both brand I can turn on or shutdown the server one per one using a browser interface (tested from local network). I'm planning to add 10 more server (probably HP because they are very cheap) and I definitively need to be able to turn on or shutdown all server at once. I'm pretty sure that lot of professional Admin have been in the same case than myself, so there should be a way to easily pilot several server at once. Thanks for your help.

Improve this question
7
  • Welcome to serverfault. Unfortunately, this site is for sysadmins only, and thus your question is off-topic
    – MichelZ
    May 22, 2014 at 5:40
  • 2
    I believe that such question may have been asked by sysadmins as well.
    – Laurent
    May 22, 2014 at 6:11
  • While this might be true, as ewwhite points out, shutting down servers is not common amongst professional sysadmins.
    – MichelZ
    May 22, 2014 at 6:13
  • May I ask why you're buying physical servers instead of renting space in the cloud? Are you dependent on GPU acceleration?
    – pauska
    May 22, 2014 at 6:24
  • 1
    of course, no secret: - using a remote farm takes hours to upload asset - some function are not supported by the commercial render farm - I mainly use it to speed up the render preview process, it's what is time consuming. By using distributed rendering I can get a very fast result to fine tune material, light, shader... - gpu do not support a lot of feature and can't be used for production at this stage
    – Laurent
    May 22, 2014 at 6:42

3 Answers 3

Reset to default
4

Why do you need to power these server all off or on at once? It's actually not a common request or action amongst professional sysadmins.

Either way, you can either use a combination of ILO and IPMI tools: Using the ILO SSH interface for the HP servers and ipmitool for the Supermicro servers.

Or you can just use ipmitool to control both system types.

In addition, both the HP and Supermicro ILO have a command line interface. That may help as well.

Improve this answer
3
  • thanks, Simply because the render farm is dedicated to my own needs and it is not needed to render 24/24 7/7. So I need to turn it only when I have to render a project and off until I start another project. How are admin dealing with thousand servers for example? Even if it's only once every 2 years, for whatever reason, they probably don't do it one per one.
    – Laurent
    May 22, 2014 at 6:09
  • We do it using the various means on the OS. We do not shutdown the server, we "restart" it, which eliminates the need for ILO/IPMI to startup the systems
    – MichelZ
    May 22, 2014 at 6:14
  • IPMIView from supermicro has the ability to control servers by group. Not sure it will be compliant with HP, I need to test it.
    – Laurent
    May 22, 2014 at 7:58
0

Can I make an alternate suggestion? (I'm going to ignore the fact that servers don't like being switched off - Perhaps you're booting from a SAN? :P)

I'm also going to assume that they're running Linux - Because rendering on Windows is terribly slow and inefficient.

How about using Wake-On-LAN to wake them up from their slumber, and using SSH to shut them down. You could script SSH quite easily, with any one of a number of tools, (I'd probably look at using Ansible... but that's how I think..)

If they are running on Windows, and you want windows to be the cluster controller, you could still use Wake-On-LAN to start them, but use Powershell Remoting (or similar PSExec runes) to shut them down again.

Or you could use ILO/IPMI/BMC tools to have control over the power supply.

There's another dirtier method, where you put them all on a network-enabled PDU, then you could just switch the port relays to OFF, and it's like unplugging the power leads.

Improve this answer
8
  • lol @ PDU suggestion :) I would definitely not recommend this, not even as a last resort...
    – MichelZ
    May 22, 2014 at 11:31
  • PDU is fine as long as you've done halt or similar first.
    – Tom O'Connor
    May 22, 2014 at 13:23
  • Yes, but how does that help? the server stops on halt anyways
    – MichelZ
    May 22, 2014 at 13:24
  • It'll stop it using standby power. So you wouldn't be able to WOL it, but assuming you've configured the BIOS correctly, you could tell it to start on power-on.
    – Tom O'Connor
    May 22, 2014 at 13:30
  • Yeah, I understand that... it just feels like you want to hard-power-off them while running, and then switch them back on :)
    – MichelZ
    May 22, 2014 at 13:32
0

Look at Ansible as an orchestration tool that can talk to your iLO interface (or alternatively Wake-on-Lan) and instruct systems to boot (or boot from a boot-media).

Ansible can also instruct the OS operating system to update, reconfigure or shut down, works for most (if not all) flavors of Linux, Unix and Windows.

For most OSes it uses SSH/python, but for Windows it takes advantage of WinRM/Powershell.

The following Ansible modules can help: hp_ilo, wakeonlan, win_reboot, etc...

Ansible doesn't just do hardware and operating system, it can also manage cloud-environments, network-equipment, applications and everything else with a proper API.

Improve this answer

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .