Description of problem:
Add EUFI support for Ironic for Ocata
Create an upstream bug & BZ for this in Ocata timeframe
Need Red Hat help to implement in Ocata
Version-Release number of selected component (if applicable):
Steps to Reproduce:
Can be split into two BZs.
Hi! We've had full UEFI support in Ironic for quite a while, and in Director since Mitaka or even LIberty (enabled by default in Newton/OSP10). Please feel free to file specific bugs if you see some missing features or encounter any issues.
*** This bug has been marked as a duplicate of bug 1093245 ***
Last time we tried it on JS HW it did not worked as expected.
Adding a few folks to cc list to dig deeper.
There were some problems with RHEL 7.2 (and early RHEL 7.3) kernel which should be fixed in upcoming RHEL 7.3. Otherwise please file bugs.
Reopening as a feature request against DRAC driver specifically.
The patch was abandoned due to Miles leaving the team.
For this one, the upstream patch  has been abandoned and therefore not likely to make OSP 12. The recommendation is to assign someone from Dell EMC team to drive this upstream, along with the QA. Red Hat engineering can offer guidance, help on the review side, and of course bringing downstream.
Sorry, also forgot to add the comments from engineering I got recently:
This BZ should verify if the pxe_drac Ironic driver can be used on nodes with UEFI enabled.
The tests are simple: register node in UEFI mode with Ironic and deploy image or just try to deploy OSP with director on a Dell node in UEFI.
Miles Gould, who recently left Red Hat, noticed that changing the boot order from the pxe_drac driver when in UEFI wasn’t possible and tried making some changes in the driver which proved unstable and he decided to abandon them (https://review.openstack.org/#/c/420107/). Those changes were just part of the tests but it could well be that UEFI in Dell also requires specific settings in the firmware.
Could Dell verify this if they haven’t done that yet?
Do you know if JetStream uses pxe_drac and works in UEFI?
JS 10 uses pxe_drac, but deliberately puts the overcloud nodes into legacy boot mode because UEFI boot mode does not work.
I know that Miles had tried getting ironic to work with the pxe_drac driver and UEFI mode, and it didn't work.
I had spoken with him at an OpenStack summit, and we agreed that Dell would push RAID configuration forward, and he would push UEFI booting forward. As a result, we've been focused on RAID config (we have quite a few changes to push upstream) and really haven't looked at UEFI booting at all other than to review his proposed patch.
Hi Chris - just wondering if you will get to this during the Queens cycle? If not we'll change the flags to push it off to rhos-14.
Hey Bob, we've started working on this, but it definitely won't make Queens. Feel free to push it out.
I submitted a change for review in Gerrit which resolves the bug -- https://review.openstack.org/#/c/545184/ . It is in the openstack/ironic project. No changes to the openstack/python-dracclient are needed.
It is marked Do Not Merge (DNM), because more automated unit tests and a release note are needed before it can be merged. Those will soon be forthcoming.
The functional code (production code) changes have been completed. They have been thoroughly tested against actual hardware configurations. It successfully lends support for UEFI boot mode to Dell EMC PowerEdge 13th and 14th generation servers with a wide variety of disk devices, including Dell PowerEdge RAID Controller (PERC) RAID volumes, Dell Boot Optimized Storage Solution (BOSS) RAID 1, Host Bus Adapter (HBA) controlled JBOD, BOSS SSD, and NVMe. Regression testing of legacy BIOS boot mode against those same configurations passed. Finally, 12th generation servers continue to work with legacy BIOS mode and fail with UEFI. The change is not intended to resolve the bug on 12 generation servers.
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, and where to find the updated
files, follow the link below.
If the solution does not work for you, open a new bug report.
Was there a backport of this to OCP13? We are experiencing similar issues on two client deployments at the moment.
(In reply to Joe Wood from comment #32)
> Was there a backport of this to OCP13? We are experiencing similar issues on
> two client deployments at the moment.
Yes, it was backported to OSP 13. It first appeared in OSP 13 Zstream 2. More details are available at https://bugzilla.redhat.com/show_bug.cgi?id=1614964. Also, the upstream Ironic change was cherry picked onto the stable/queens branch -- https://review.openstack.org/#/c/588843