We forgot to revert it. This job covers software RAID and manual
cleaning, so it's very important to avoid regressions, even if it costs
us some rechecks from time to time.
Change-Id: I2446afeaca866ffc3131b5e9f266526f35fc5ed7
It seems like tinyIPA silently replaces /dev/md/esp with /dev/md127.
Find the next free /dev/md device and use it instead.
Also rescan the resulting device before copying files.
Change-Id: Ie04f530be434c4b1561e75f387b9da679e4607e0
Depends-On: https://review.opendev.org/c/openstack/ironic/+/827129/
Move the software RAID code path from grub2-install to
efibootmgr:
- remove the UEFI efibootmgr exception for software RAID
- create and populate the ESPs on the holder disks
- update the NVRAM with all ESPs (the component devices
of the ESP mirror, use unique labels to avoid unintentional
deduplication of entries in the NVRAM)
Story: #2009794
Change-Id: I7ed34e595215194a589c2f1cd0b39ff0336da8f1
Pair nodes dynamically via a distributed coordination backend for
network burn-in. The algorithm uses a group to pair nodes: after
acquiring a lock, a first node joins the group, releases the lock,
waits for a second node, then they both leave, and release the lock
for the next pair.
Story: #2007523
Task: #42796
Change-Id: I572093b144bc90a49cd76929c7e8685ed45d9f6e
We have updated the yoga testing runtime to keep the
py36 testing.
- https://review.opendev.org/c/openstack/governance/+/820195
Unit tests job template is also updated to keep python
3.6 as a voting job. So with the py3.6 and py3.9 testing as voting
job template, we are keeping python 3.6, 3.7, 3.8, and 3.8 as
tested versions in the Yoga cycle.
- https://review.opendev.org/c/openstack/openstack-zuul-jobs/+/820286
This commit re-add the python 3.6/3.7 versions in setup.cfg classifier.
Change-Id: I0f03a7f5bb2aa07c2ec2aab1a8ebfddc0c70ca87
In order to ease logging of the various burn-in steps, this patch
proposes options to define the outpout files for all burn-in steps:
{'agent_burnin_cpu', 'agent_burnin_vm', 'agent_burnin_fio_network',
'agent_burnin_fio_disk'}_outputfile via a node's driver-info.
Story: #2007523
Task: #44102
Change-Id: I327cae5949d38e738d3c535487b3795d00ad8f1e
Doing this will cause it not to zero out the entire
block device which can be very costly on a slow HDD.
Story: 2009227
Task: 43315
Change-Id: I62ba2afc037d9844387e6b0984fe5008779d95d2
Add the option to run a SMART self test right after
the disk burn-in. The disk burn-in step will fail if
the SMART test on any of the disk fails.
Story: #2007523
Task: #43383
Change-Id: I1312d5b71bedd044581a136af0b4c43769d21877
As discussed during the ironic meeting on 29/11/2021, the community has
decided to continue testing the code against Python 3.6 to keep
compatibility with distributions that have native support for it, such
as CentOS Stream 8, Ubuntu Bionic, Debian 10, and openSuse Leap, at
least until it's possible and reasonable considering the support granted
from other dependencies and libraries, and not beyond the Yoga cycle.
Change-Id: I97ad7c4acdb278e2849a598c5a02a71fc67cc81a
This commit changes the lambda usage in the RealFilePartitioningTestCase
to autospec to avoid problems with unexpected args.
Change-Id: I21356a7783f105dde9ff0d3777e2a06f3f28a786
This function returns the complete block device record, not just number.
Fixes regression in 89bc73aa0105850c6ae44428642e31802bba3b20.
Also fix the incorrect job in the gate queue, which prevented us from
catching this issue on merging.
Change-Id: I4cbc359ceabfc193ce18fed14a1952359460e7d9
Replace the execute wrapper from utils with execute from ironic-lib in
hardware.py
Adjust unit tests as needed.
Change-Id: I63a3b0407b2ca2246bd0e6624bfa0f748c0d73f7
This call is very useful for custom deploy implementations, such as one
we maintain for OpenShift. Splitting it out also makes image.py slightly
more manageable.
The get_partition call is moved to partition_utils.
Change-Id: I60a6a2823d3eb27a4ae78e913e3655dae7b54ffe
We use basically the same function in two modules in the same way, let's
put that in a common place.
Change-Id: I4016e43f2cb102d4327bafcc8a2f90112a6f944a
The caller knows what step it invokes, there is no point in repeating
it in the error message. There is also no need to wrap the exception
if it's a RESTError or an ironic-lib exception already since they
are normally detailed enough.
Only leave a detailed message when an unexpected exception happens.
Change-Id: I1d8ca1e7ed1462159e4ae5f0bcf58686f6a2681c
Use add instead of update to re-read the partition table with partx.
See [1] for more details.
Co-authored-by: Arne Wiebalck <arne.wiebalck@cern.ch>
[1] https: //opendev.org/openstack/ironic-python-agent/commit/dc8c1f16f9a00e2bff21612d1a9cf0ea0f3addf0
Change-Id: I2336e22dadc790cfbde87904612fcaa3b8c501db
Re-read the partition table with 'partx -a', rather than 'partx -u'.
This should fix an timing issue where the bootloader installation
fails to mount the EFI partition from a whole disk image since it
is not yet aware of the new partitions (observed with both, the
iscsi and the direct deploy interface).
Change-Id: If5da3075e813ae01df3decf8f0647aba111b0515
I accidently put colons on the test data and remembered taking the
colon character out of the regex I was working on, but apparently
left it in, and accounted for the active entry indicator flag
which appears to have inconsistent support across vendors.
The regex has been fixed, and a test added from a Lenovo SR650
which has some additional string entry data in the UEFI output
which may separate entries.
Change-Id: I1f67b0fb1f645fa82e98bd7c7bba3ffc7755cc74
Some firmware seems to take an objection with EFI nvram
entries being deleted after one is added, resulting in the
entire entry table being reset to the last known good state.
This is problematic, as ultimately deployments can time out
if we previously booted with Networking, and the machine, while
commanded to do other wise, reboots back to networking regardless.
We will now delete entries first, before proceeding.
Additionally, for general use, this pattern may serve the
community better by avoiding cases where we would have
previously just relied upon efibootmgr[0] to warn us of duplicate
entries.
[0]: 103aa22ece/src/efibootmgr.c (L228)
Change-Id: Ib61a7100a059e79a8b0901fd8f46b9bc41d657dc
Story: 2009649
Task: 43808
Even if journald is present, there is no guarantee that IPA logs there
(this is the case in container-based ramdisks).
Change-Id: Iceeab0010827728711e19e5b031ccac55fe1efde
Based on unit tests, this was done intentionally, but I don't see
reasons for that. It makes refactoring much harder, because sometimes
you need to mock both execute functions and test them separately.
In the end, utils.execute should be removed.
Change-Id: I5a9c694ebe626c54f219d4870eab0a592777518d
* Use the same TLS parameters as everything else
* Respect image_download_connection_timeout
* Do not ignore HTTP errors
Change-Id: I84f8021f731186d82e44ac3d4ef2d12df13f830a
The EFI partition UUID may be None and this will break
the fstab editing. While this is not necessarily fatal when
instantiating a node, it creates an exception at the end of
bootloader installation, so only attempt to add a line to
fstab when the UUID is not None.
Change-Id: I68799980e67c05afe4ca68ca9733605dd166d54d
This patch fixes a race during software RAID creation:
we create the partition with parted, the kernel then
notifies udev, but we need to wait for udevd to create
the device files before calling mdadm to create the
md device.
Credits to jcosmao for finding this.
Change-Id: I642f28acc351cf50263e37dfbc8468bf59de2cc5
