UPS problem finally solved for vSphere

The challenge

Everybody will agree on power outages too happen more likely due to more and more decentralized sustainable energy sources being used.

Therefore, UPS systems have been and will be a mandatory detail in every virtual infrastructure.

The problem

Most of the UPS provide shutdown mechanism said to be specifically designed for vSphere. Unfortunately, they never work in a manner that you will expect. Not only do they mostly rely on pre-configured scrips that sometimes even contain plaintext password. Even more they just trigger the shutdown of your affected ESX host and therefore just shutdown your machines in a random order. Of course, you could define a specific order but once you vMotion your VMs around the order gets lost.

Risk of data corruption

Shutting down your workloads in the wrong order can not only affect your shutdown time negatively. It is on top of that a risk for your data. E.g. an Exchange server needs at a least one domain controller in reach to shut down properly. If none is found the process just gets stuck and the server keeps waiting in the process. Once the UPS battery is empty this server will crash in the middle of its shutdown. Database corruption is likely to happen.

The solution

What we need is a cluster-wide shutdown order that does survive vMotion/HA/DRS activities.

hypnOS is here to fill that gap.

Installation and configuration is simple: First add your UPS system(s):

Connection to the UPS boxes is realized via SNMP to have a reliable standard for a wide range of UPS vendors.

Once you have the UPS systems complete add your ESX hosts and tell hypnOS which UPS backs which ESX host.

All VMs on each host are automatically added to the system and will always be identified by their unique ID within vSphere. Even if they move between hosts.

Finally define your shutdown order graphically by using drag and drop. Three shutdown tiers allow a fast categorization.

Optionally you could fine tune the order within each tier. In our previous example the Exchange server would go to the first tier while the domain controller should be the last man standing in the third tier.

Now hypnOS will be waiting for the next power outage to occur.

In contrast to the UPS vendor provided solutions your will know if the system is functional. Once hypnOS loses its connection to a host or UPS system you will be informed. So, password changes will be recognized immediately.

Hands-on videos

For detailed information and a hands-on demo watch these videos provided by idicos, the makers of hypnOS.

The principle behind the software

Hands-on shutdown demo

Whats next

The upcoming version 3.0 will also include fully automatic migration of certain VMs from the affected hosts to still powered host in other zones while others might be shut down even there to give room. Stay tuned…

Leave a Reply

Your email address will not be published. Required fields are marked *

Solve : *
10 + 21 =