We are running several CHR instances on our virtual environment but lately we have seen some peculiarities.
For example, we were running one CHR on our VMWare environment. This VMWare environment uses hosts with the following CPUs: Intel Xeon E5-2660 v3 2.6GHz
The CHR had 2 sockets of 1 core, 2GB RAM and 1GB hard disk. This CHR had one network interface (VMXNET3). There was about 400Mpbs throughput on this CHR and the CHR was running around ±60% load.
We tried migrating this CHR to our new VMware environment which is running on AMD EPYC 7551P 32-Core Processor. The migration was executed by shutting down the VM, migrating to other host and then powering on the VM.
We then had a lot of packet loss as well as messages there was a loop (while there was definitely no loop; the same setup/configuration had been running on VMWare for over a year.).
After a while we decided to rebuild the VM on our old Xen environment (which uses the same Intel Xeon E5-2660 v3 2.6GHz CPU architecture as our old VMWare environment).
The packet loss messages were gone and the CHR was running smoothly again. There was again around ±400Mpbs on the CHR but this time the CHR was running at around ±6% CPU load.
So my question is 2-fold:
- Why is the difference in CPU load so big between Xen and VMWare? Whilst the VM has the same resources and same throughput.
- What could be the reason for the loop problems on the AMD environment? Whilst the network and configuration remains exactly the same.
Kind regards
Frédéric Saelens
Citymesh