Start a Conversation

Unsolved

This post is more than 5 years old

T

99911

November 10th, 2012 10:00

What Dell UPS managemet software setup to use for controlled shutdown?

I've been looking through user guides, white papers, etc. but I am still somewhat unclear about what would be my best course of action for ensuring a controlled and properly sequenced shutdown of our servers in case of power outages (we just had one).

It's a comparatively simple and I would guess very common setup:

One rack with 2 PDUs, and:

One Dell 3750W UPS with an EBM

8 physical servers (3 x R510, 2 x R310, 2 x R620, 1 x 2950 III)

3 switches (2 x 6224, 1 x 5548)

1 Equallogic PS4100X SAN

2 of the R510's are HyperV hosts (in a cluster) hosting about 10 VMs. We use SCVMM to manage the VM's. Right now the SCVMM server isn't actually powered by the UPS, but we will remedy that.

What I am looking for is simply a way to perform a controlled shutdown of the physical machines AND of the HyperV VMs in the appropriate order, taking account of various dependencies. These are not just the "basic" dependencies of servers depending on the SAN and the switches, but also dependencies between individual servers - physical and
virtual (for example: Our main file server is virtual).

I have already mapped out all of the dependencies, and I know the shutdown sequence I want, but I do not see how to implement it in either ULMN or MUMC.

For the record: I am not interested in automating SCVMM Live migration since in the event of power failure, there will not be any place to migrate to.

I am looking for the simplest possible way to achieve my goals.

All of the servers (physical and virtual) run Windows Server 2008 R2 (except the 2950, which runs "R1").

Right now, all I have set up is the UPS network connection and the management web site hosted there.

Some stats from the UPS:

When at 100% capacity there is usually about 25 minutes remaining time. The load is usually about 50 %. The AC output is about 1.8 kVA.

I suspect the EBM is not used as all we did when we set it up was to connect the cables. I am looking into that on Monday:

The point of this is that there will likely be plenty of time for a sequenced shutdown.

Update: Having properly attached the EBM, the UPS now reports 50 minutes remaining time.

44 Posts

July 3rd, 2013 07:00

Hi Thomas,

Although it's a (very) late answer I thought I'd provide one for others who might be asking similar questions in the future.  There are two approaches to managing a virtual infrastructure with the Dell UPS Management Software:

1. Agentless (VMware with vCentre, Xen with XenCentre):  You can just install the Multi-UPS Management Console on a Windows VM (or management machine), add the controller credentials and manage the host response directly from the VMWare/Xen plugin.  A good guide for VMware is here:

http://www.dellups.com/pdf/Dell-TIN-UPS003_VMware_vCenter_Shutdown_Scenarios-10-11-12.pdf

And a video of the software in action can be seen on www.dellups.com/software/en

2. Agent Managed (traditional OS, HyperV):  Here you need to install the UPS Local Node Manager (ULNM) agent on each OS/host and manage them either through their own interface or centrally for multiple agents through the MUMC interface.  A general guide to setting this up is available on this link:

http://www.dellups.com/pdf/Dell-TIN-UPS002_Common_Software_Installation_Scenarios-10-11-12.pdf

As a final point to note, agentless management does rely on the fact that both the MUMC controller, plus the host controller are in constant communication with hosts.  If the hosts are located remotely from the controller (e.g. off-site) and there is a risk to network interruption, then it may be more appropriate to use the ULNM agent on each host instead.   This way, even if the controllers lose contact with the hosts, the ULNM agent can still perform shutdown if needed. 

Hope this is helpful.

Andy

44 Posts

July 4th, 2013 09:00

Andy,

Thanks replying. Better late than never :-)

We are a HyperV shop, so I guess it's option 2 for us.

Did you mean that we should install ULNM in each vm or on each host?

And are you saying that we can sequence all the shutdowns or vms and hosts from Mumc?

 

I read through the mumc user guide but didn't see anything on how to do that.

I also see that I would need to integrate a python script for shutting down our SAN. can I launch that from mumc?

44 Posts

July 4th, 2013 10:00

Hi Thomas,

Glad I can still be of some use!  :-)

The best doc is the "Common Scenarios" one linked in my previous post, it's more linear in approach and only tells you what you need to know.  The UG's are quite extensive so easy to get lost.

You install the ULNM agent on each Hyper-V host, the VMs do not need an agent (phew that will save some time!).  You can then configure the ULNM agents directly through MUMC - add them in the Auto Discovery Menu by doing a range scan for the IP addreses of the hosts running the ULNM agents, then edit the settings in the "Node Settings" menu.   You can also make any UPS settings you want here too.   MUMC can be run on a Windows VM or a physical machine, it doesn't matter.

Finally you need to make some settings in the Hyper-V manager to tell it what to do with the VMs in the event that the OS receives a shutdown instruction.  ULNM makes this shutdown instruction if the UPS runs on battery for the prescribed period of time.  The effect is similar to clicking Start -> Shutdown.

If the UPS runs on battery long enough to initiate shutdown, the Hyper-V service controlls the VM response and does whatever it is instructed to do before allowing the host to safely shutdown.  

ULNM can launch either a shutdown event or a script, but not both at the same time (at last I'm 99% sure on this one).  So, leave the ULNM to shut down the Hyper-V agent and instead, use the MUMC console to manage initiation of a script.  Within MUMC you can go to System -> Modules Settings -> Edit Modules Settings.  Enable "Shutdown".  This will allow MUMC to do management of ULNM agents but also now enables MUMC to act as a shutdown agent in its own right.  It creates a new system menu on the left called "Shutdown".  Then in the shutdown settings, choose script.  Now you can launch any kind of script you like.

I'm not sure what SAN you're using, but below is an example script of how to handle a Dell EquaLogic system.  Maybe you'll find this useful, i hope so.  This can be saved in a .ps1 file and run from the UPS management software when needed.

Hope this is more useful information!!!!

Andy

# --cc @ 2012/04

# Shutting down EQL arrays

# DISCLAIMER HERE ....

# Customizing and executing this script is entirely at your own risk.

# Shutting down members is disruptive for initiators as it will take the volumes offline, so please make sure initiators are disconnected first.

# some vars for consistency

$myUser="grpadmin"

$myPwd="grpadmin"

$myGrpIP="172.28.12.120"

$myTestPool="maintenance"

$myTestMember="Aillen"

# Make sure we're having the PSTools loaded

Import-Module 'C:\Program Files\EqualLogic\bin\EqlPSTools.dll'

# Connecting to group over MgmtIP

Connect-EqlGroup -GroupAddress $myGrpIP -UserName $myUser -Password $myPwd

# Shutdown selective member

# to use this method, uncomment line below and adjust var parameters

# Stop-EqlMember -MemberName $myTestMember -Force

# Shutdown selective pool

# to use this method, uncomment line below and adjust var parameters

# Get-EqlMember | Where-Object {$_."StoragePoolName" -match ("$myTestPool")} | Select-Object "membername" | Stop-EqlMember -Force

# Shutdown ALL members in the group

# IMHO, this should only fit the bill for very small deployments with all members in the default pool where you just don't mind shutting them all down regardless...

# to use this method, uncomment line below

# Get-EqlMember | Select-Object "membername" | Stop-EqlMember -Force

# DEMO RUN

$demo = Get-EqlMember | Where-Object {$_."StoragePoolName" -match ("$myTestPool")} | Select-Object "membername"

Write-Host "I should have send shutdown commands to :`t" $demo

No Events found!

Top