AVD-Multiple hosts stuck at 100% CPU with no disk or network usage after hibernation
We have a few thousand D4SV5 AVDs with hibernation currently in production that we've been using since late 2024.
Since around April we've been getting a bunch of users calling into our Help Desk unable to connect and a simple stop start fixes it.
Pulled some data and besides for allocation issues (spread out through all subscriptions) we have a bunch of users getting RDS unhealthy errors.
Exported all of Event Viewer for a host (all users have personal hosts) and found that the logs are empty from around 5AM until 9AM (service desk did the stop start around then) when it showed that the last shutdown was unexpected with no crash files. We also have other monitoring agents on these hosts and they showed nothing.
However according to Azure, it started up at 6AM (scaling plan wake up), reported accelerated networking issues (required for some reason on a D4SV5) at 6:28AM and the user got the RDS error at 8:32AM.
Most notably, in the performance charts on Azure it shows from 6AM to 9AM the CPU spiked at 100% as the first data point until the end with absolutely no network or disk traffic or use.
Azure VM Team is insistent that this is a Guest OS problem.
How is that possible?
From all my years of Windows and VM experience that seems like a Hypervisor that assigned CPU but couldn't assign memory or disk.