This issue affects Linux images based on Microsoft Azure NV-series and NVv2-series GPU-optimized virtual machines with licensing provided by Microsoft Azure. If the runlevel in such an image is changed, the image fails to acquire an NVIDIA virtual GPU software license.
This issue occurs because the hypervkvpd daemon stops if the runlevel is changed. The hypervkvpd daemon is required for communications between the guest virtual machine and the host. When the daemon is no longer running, the nvidia-gridd service fails to acquire a license from the host.
When this issue occurs, error messages similar to the following messages are logged:
Dec 24 08:08:24 nvdavgpugrhel nvidia-gridd: Started (12402) ... Dec 24 08:18:28 nvdavgpugrhel nvidia-gridd: Maximum number of retry attempts reached Dec 24 08:18:28 nvdavgpugrhel nvidia-gridd: Failed to acquire license (1)
This workaround requires administrator privileges.
Start the hypervkvpd daemon.
# service hypervkvpd start
Continue with the installation of the NVIDIA graphics driver or restart the nvidia-gridd service.
Not an NVIDIA bug
Affected Products and Services
- NVIDIA virtual GPU software
- Microsoft Azure NV-series and NVv2-series GPU-optimized virtual machines with licensing provided by Microsoft Azure
- Red Hat Enterprise Linux 7.x