Adding PDU support #27

ErwanAliasr1 · 2024-07-09T14:35:32Z

This pull request is about adding the PDU support to hwbench & hwgraph.

This first version supports Raritan PDUs but can be extended to other brands.

hwbench/environment/vendors/bmc.py

hwbench/environment/vendors/monitoring_device.py

hwbench/environment/vendors/bmc.py

documentation/monitoring.md

hwbench/bench/monitoring.py

Custom vendors inherit from Vendor class. If the default init is override, it must call the parent class init to ensure all vendors behave the same. Signed-off-by: Erwan Velu <e.velu@criteo.com>

hwbench/environment/vendors/monitoring_device.py

The monitoring config file is fully dedicated to the BMC. This commit is about adding a "type" to each section to let room for future monitoring components like PDU, CDU etc.. This commit is adding "type=BMC" for the current BMC examples. The BMC code is also moved out of vendor.py to make it easier to read. Signed-off-by: Erwan Velu <e.velu@criteo.com>

Future work to add different type of monitoring devices like PDUs, CDUs, etc.. requires a generic class to share some behavior. This commit is moving parts of the BMC class into the generic MonitoringDevice one. It also creates a new find_sections() function to make the lookup of given type of MonitoringDevice inside the configuration file. Signed-off-by: Erwan Velu <e.velu@criteo.com>

The current code is using get_ip() to retrieve BMC IP. It uses ipmitool or ilorest to detect the BMC ip address. This is a limitation since we possibly want to force the url instead of an autodetect. This commit is : - renaming get_ip() to get_url() All drivers must return the BMC url and not its IP address - ensure that BMC.get_url() returns an https:// string instead of an IP addr - removes the "https://" append in connect_redfish() A test is now implemented to ensure drivers properly return a valid get_url() - allow url to be defined in the monitoring.cfg file This gives more flexibility for future usage. - add a mocked monitoring.cfg file A very simple configuration file is added to get the minimal BMC structure. Signed-off-by: Erwan Velu <e.velu@criteo.com>

This commit is about graphing PDU metrics. The commit is very light since it only requires adding PDU as a new source of power metrics and plot some specific graphs. The code is very light at PDUs are just an additional source of data in the existing data set. Signed-off-by: Erwan Velu <e.velu@criteo.com>

This commit add PDU as a type of monitoring device. A monitoring.md documentation is added to detail how the monitoring works and the way to use it. PDU monitoring is optional and needs to be defined inside the monitoring config file. It is possible to setup multiple PDUs configuration in the same file. This version implements a first driver for Raritan PDUs with Redfish API. At statup time the output looks like the following: Monitoring/PDU: initialize myPDUgroup with Raritan driver @ https://1url.pl/ Model: 'PX3-5722V-V2' FW: '4.2.0.5-50274' Serial: 'XXXXXXXXX' Monitoring/PDU: initialize myPDUdirect with Raritan driver @ https://url.pdu/ Model: 'PX3-5722V-V2' FW: '4.2.0.5-50274' Serial: 'XXXXXXXXX' Monitoring/BMC: Thermal metrics:1xCPU, 1xIntake Monitoring/BMC: Fans metrics:10xFan Monitoring/BMC: PowerConsumption metrics:65xCPU, 4xBMC Monitoring/BMC: PowerSupplies metrics:2xBMC Monitoring/PDU: PowerConsumption metrics:65xCPU, 4xBMC, 2xPDU On the implementation side, the vendor drivers, like Raritan here, are inheriting the PDU class and only override methods if necessary. Signed-off-by: Erwan Velu <e.velu@criteo.com>

anisse

LGTM

ErwanAliasr1 force-pushed the pdu branch 5 times, most recently from 175fbca to ab30e99 Compare July 10, 2024 14:45

ErwanAliasr1 requested review from anisse and beorn- July 10, 2024 15:13