PowerChute 4.2.0 - managing an advanced, non-redundant setup for vCenter with 4 Hosts

[Anonymous user]

Greetings,

after looking all over the place (and being none the wiser), I come here in hope of specific tips on the following problem:

Our server environment consists of 4 VMware hosts managed via a vCenter virtual appliance. Hosts 1+3 and 2+4 form pairs (i.e. being in the same rack), with each pair sharing a Smart-UPS SRT 6000, and with a total of 2 UPS. The UPS are thus not redundant. This is due to the host pairs being divided into separate server rooms within the building. Both UPS are managed via PCNS 4.2.0, which is installed on a standalone non-virtualized system. The storage environment is protected separately.

We have used the manual to set this up as an advanced UPS configuration. Each pair of VMware hosts is assigned to the correct outlet group on the respective UPS in the host protection menu.

After configuring the settings for VM prioritization and shutdown durations, we enabled shutdown for the "UPS On Battery" event. When simulating a power failure on one of the host pairs, however, all hosts went into power failure state, and consequently began to enter maintenance mode. In this scenario, we had deactivated DRS, as that is another problem entirely.

Thus my main question: is there a way to manage our UPS via PCNS in such a way that it can determine which server is now running on battery, and thus only start the maintenance mode sequence on these hosts? I should think this to be possible, but I know not how.

I have deliberated on the following options:
- two separate installations of PCNS, each managing one pair of VMware hosts and one UPS - but this would make use of the plugin impossible, and perhaps pose problems when using DRS later on
- adding the VMware hosts separately, not as vCenter cluster - but this would make VM prioritization impossible as well as a (later) use of DRS

Is there a way to make our intended setup work, or will we have to reconsider? I can provide a graphical oversight if my explanations above are insufficient. Thanks for any tips you can give us!

[BillP]

Hi,

Where is vCenter Server running? 

Have you review Application Note 180? 

[Anonymous user]

Hi there,

thank you for replying. First of all, for better clarity, I have attached a visual representation of our current setup. Our vCenter is running as a virtual Appliance on the protected host cluster. Note that we chose to protect the cluster instead of each individual host. Referencing application note 180, our setup is rather similar to the one on page 6. If I were to describe our differences, I'd say ESXi host A would be our hosts 1+3 and host B would be hosts 2+4. Also, it deviates from that setup by having separate main fuses for Hosts 1+3 and Hosts 2+4 as well as the UPS systems - but I'm not sure this is of importance.

In Scenario D of application note 180, there is a marked difference in behaviour of maintenance mode and shutdown compared to our current setup. Deviating from the given scenario D, our scenario looks like this:

1. PowerChute reports that the UPS powering ESXi Hosts 1 and 3 is on battery.
2. Shutdown delay for the On Battery even elapses. PowerChute starts a maintenance mode task on all Hosts (1-4).
3. PowerChute starts to shut down VMs as per the prioritization list.

This is not the behaviour described in the example scenario, and I can't wrap my head around why it behaves this way. We even deactivated VM migration for testing purposes, as you can imagine 4 hosts going into maintenance mode and trying to migrate towards the other hosts didn't go well at all.

I have also attached a screenshot of our current host protection settings. I'm not too certain what I'm missing at the moment, so any hint is welcome!

[BillP]

Hi,

With advanced redundant UPS configuration for VMware you can set the individual UPS to power down in the event of a power failure or you can have all UPS power down. When running PowerChute Setup you can set each UPS to act in a specific manor. In the screenshot below I set the PCNS UPS to execute virtualization shutdown sequence which will cause all of the hosts to be powered down even though they are on different UPS.

Since UPS Host .23 is only running VMs (not VCSA) I have configured it to not execute virtualization shutdown sequence since I do not want the other 2 UPS powered down if this is the only UPS on battery. I did configure migration of the VMs on this host so they will be moved to the other host .22 if this host is going down. 

For host .22 I configured execute virtualization shutdown sequence since it is running VCSA and I don't want that VM migrated if the UPS goes on battery. 

[Anonymous user]

Thanks again for replying!

If I understand you correctly, I would have to bind the VCSA to one half of my cluster's hosts, which sounds reasonable. However (as I understand it), in the suggested setup, host .22 going on battery would still shut down the entire cluster?

[BillP · Accepted solution]

Hi,

Yes. If .22 goes on battery host .23 will be powered down. The reasoning is .22 hosts vCenter and that controls .23 so .23 should be powered down to prevent a crash due to the lack of communications with vCenter. I don't want the vCenter VM migrated because that could cause lost comm with PowerChute hence the need to power down both hosts.