f53ccb647f
So, there are cases where say you may have multiple DPUs in a physical server, each card when fully operating can consume 100-150 watts. In some cases, these cards can have external power supplies, but need the physical host in a running power-on state in order for the device to be powered on. Conversely, we also now power off the child nodes if the parent node has been requested to be powered off, since we *really* don't want cause inadvertent harm to the child node. This is realistically a fix we should backport once we sort through the details, if we agree this makes sense to do, as is. Change-Id: Ib2bfe04cdaa82264ba8bb1e71477899bb6268179
12 lines
574 B
YAML
12 lines
574 B
YAML
---
|
|
fixes:
|
|
- |
|
|
Fixes the power handling flow as it relates to ``child nodes``, i.e.
|
|
bare metal nodes which have a ``parent_node`` set, such that power is
|
|
turned off on those nodes when the parent node is powered off, and that
|
|
power is turned on for the parent node when the child node is explicitly
|
|
requested to be in a ``power on`` state. This does not apply if the child
|
|
node device has a dedicated power supply, as indicated through a
|
|
``driver_info`` parameter named ``has_dedicated_power_supply`` which
|
|
can be set to a value of "true".
|