It would also be useful to enable debug logging on the ironic-conductor log. This will show the actual request and response being made in the PATCH call. Debug logging is enabled with /etc/ironic/ironic.conf [DEFAULT] debug = True
This is the change which removed the Content-Type header[1] is unrelated and didn't explain why it was deleted. It is recommended to always set a Content-Type on requests, I'll propose adding it back upstream and I'll start the process of backporting it to 16.1 and 16.2, but if you need this change in a short time frame you'll likely need to ask for a hotfix. [1] https://review.opendev.org/c/openstack/sushy/+/607809
Steve, Is there a way that we can confirm that this will fix the issue? After I posted the output, I realized that it starts with the word "error". Is this what we expect? {"error":{"@Message.ExtendedInfo":[{"MessageId":"Base.0.10.Success"}],"code":"iLO.0.10.ExtendedInfo","message":"See @Message.ExtendedInfo for more information."}} ^^^^^ ||||| Is this normal?
Its unusual to package a success message inside an error attribute, but I've traced the code calls from ironic to sushy and nothing actually parses the content of the response, it is assumed to succeed if it returns a 200 status code.
Steve, This has all been incredibly helpful. Thanks!
Steve, Weren't we already getting a 200 status code in comment #8? The additional header changed the output, but the status code was the same. I suppose what we really need to know is if the ILO board actually *processed* the command, not really what status code it gives. Am I way off base here?
comment #8 included the Content-Type and the response was a success. The ironic packaged in 16.1 is not setting the Content-Type header in any calls, and it looks like this is why Gen9 servers are returning an error. It would still be useful to get the ironic-conductor debug logs to clearly show the exact request and response.
OK, I think I see what is happening now. In 13.x (queens) Ironic had no concept of setting the boot mode (uefi vs legacy bios). In 16.x this is now supported and the redfish driver will send a request to set the boot mode regardless of whether the redfish BMC supports that. It looks like HPE Gen9 doesn't support setting the boot mode through redfish. There is a mechanism in Ironic to silently ignore a failure to set the boot mode when the Ironic *driver* doesn't support it, but this isn't triggered when the BMC lacks support. Let me talk to my team about the best way to handle this.
The upstream fix is now merged, I can now proceed with backports to 16.1
The python-sushy changes are now in built package python-sushy-2.0.3-1.20210406233537.0241cd9.el8
*** Bug 1893927 has been marked as a duplicate of this bug. ***
Since the problem described in this bug report should be resolved in a recent advisory, it has been closed with a resolution of ERRATA. For information on the advisory (Red Hat OpenStack Platform 16.1.6 bug fix and enhancement advisory), and where to find the updated files, follow the link below. If the solution does not work for you, open a new bug report. https://access.redhat.com/errata/RHBA-2021:2097
The needinfo request[s] on this closed bug have been removed as they have been unresolved for 365 days