This post will not be the usual type of posts I write where I state a problem and then a solution but rather a question I have about the best practice for memory testing prior to deploying an ESX or ESXi host to production. As per the Performance Best Practices for VMware vSphere 4.0 guide: http://www.vmware.com/pdf/Perf_Best_Practices_vSphere4.0.pdf, VMware recommends that we:
Test system memory for 72 hours, checking for hardware errors.
I’ve recently been a part of a Cisco UCS deployment for multiple datacenters and had to run memtest on some B230 M1 blades that had 128GB of memory and noticed that even after 40 hours, memtest has only ran 2 passes:
What’s also interesting is that even with a blade that only had 32GB, 40 hours was only 3 passes on the memory:
I’ve asked my practice lead about this and he tells me that some of these guidelines may be out of date so adjust them accordingly. This leads me to wonder what other professionals are doing out there so if you’re one of them, feel free to comment.