Storage issues caused due to VM stunning

In this article, let’s talk about what happens during a backup and why backups performed during heavy transactional timeframes can cause issues.

During a backup, there are snapshots of VM’s taken.  For a snapshot, a VM needs to be stunned. (To understand VM stunning, read this great article by Cormac Hogan – https://cormachogan.com/2015/04/28/when-and-why-do-we-stun-a-virtual-machine/)

To understand the steps taken during a VM snapshot, please read the knowledge base article below –

https://kb.vmware.com/s/article/1015180

Now, we have understood VM stun and the steps for VM snapshots, and how the delta VMDK files are merged into a single disk. For a highly transactional application like databases (in this case the Oracle DB), side effects can appear due to VM stun.

During the snapshot, VMware will create a Delta VMDK file. All the new write operations will be directed towards this new Delta VMDK and will expand as the primary VMDK file is being backed up. Once the primary VMDK file is backed up, the Delta VMDK file will need to be merged with the primary VMDK file.

This is where a highly transactional application like a database can have an issue. If the I/O on the delta VMDK is higher than the rate of consolidation with the primary VMDK, the application will timeout which could cause application timeouts.

Uila is designed to catch storage anomalies that occur within your environment. Uila can easily identify unusual storage activities in your environment that cause high latencies and outages such as this VM Stun issue. With an intuitive UI like Uila, it’s impossible to miss any storage issue.