Resolved Issues#
In previous release, the following issues were identified as known issues; they have been resolved.
DGX Station A100 Fails to Boot After Applying MIG Configurations#
Issue
When MIG configurations are applied to a DGX station A100 system running DGX EL9-25.04 with
the Release 570 GPU Driver installed, the system might fail to boot when you run the sudo reboot command.
Resetting the GPUs by performing a DC power cycle does not recover the system.
Workaround
NVIDIA DGX Software for Red Hat Enterprise Linux 9 releases do not support the 570 release of the GPU Driver on DGX Station A100 systems.
To resolve the boot failure, reinstall DGX EL9-25.04, but do not set up MIG configurations. Alternatively, for systems other than the DGX B200, install a previous version of the GPU driver.
Status
Resolved in in GPU driver 570.117+ and 575.20+.
Tuned profiles do not take effect in graphical mode#
This issue only occurred on DGX Station systems and has been resolved. For instance, for the DGX Station
A100, the dgxstation-a100-tuned-profile package that is installed by the Install DGX Station A100
Configurations step in Installing the DGX Software has been modified to mask the tuned-ppd.service or
the power-profiles-daemon.service, whichever is in use, to ensure that TuneD service will be enabled after
a reboot. See RedHat Solution 6955289 for more information.
Note
For reference, the issue that was resolved is shown below. It is no longer necessary to
manually run one of the systemctl mask commands, since the appropriate dgxstation-*-tuned-profile
package now runs the appropriate systemctl mask command.
Issue
DGX tuned profiles might not take effect due to a known Red Hat Enterprise Linux 9 issue. This affects systems that use a graphical target mode.
Workaround
This issue can be fixed by running one the following commands, depending on which service, tuned-ppd.service or power-profiles-daemon.service, is in use:
Run one of these commands, depending on which service is in use:
systemctl mask tuned-ppd
systemctl mask power-profiles-daemon
Reboot the system.
reboot
Status
Resolved.