[xcpng9] port XCP-ng patches to 11.1.8 9.0 - #56
Draft
ydirson wants to merge 30 commits into
Draft
Conversation
Covers:
1. repo_gpgcheck:
a. wrong system clock putting gpg key creation in the future, causing a
yum crash (nothing special happens if the date of the signature is in
the future ¯\_(ツ)_/¯)
b. other yum crashes due to uncaught gpg exceptions (if any)
c. lack of repomd signature (while repo_gpgcheck is in force)
d. signature done by other key than the one in ISO ("repomd.xml signature
could not be verified" ¯\_(ツ)_/¯)
2. gpgcheck:
a. RPM signed with unknown key
b. unsigned RPM referenced by unsigned repomd (no-repo-gpgcheck)
c. RPM re-signed with unknown key, unsigned repomd (no-repo-gpgcheck)
d. RPM overwritten with another RPM signed with known key (diagnosed
through hash but, same diag as 2.c)
e. delsigned/resigned/etc RPM, unchanged repomd (same diag as 2.c/d)
Does not cover notably:
- unsigned RPM referenced by (re)signed repomd
In some cases Yum does not give an error, but dies because of an
uncaught exception, which makes this check quite brittle, but in the
worst case if messages change, we still fallback to the original
"Error installing packages" message.
Signed-off-by: Yann Dirson <yann.dirson@vates.fr>
XAPI now rejects the default keysize of 7.x era, which must be regenerated before upgrading to 8.3. Let the installer refuse to initiate a situation where a Rolling Pool Upgrade would be unable to proceed, with not-yet-updated slaves holding the running VMs getting refused connection to the updated part of the pool. To be extra-safe, add a check that the key is indeed a RSA one as that's all current XAPI supports, this will make sure this code gets updated when XAPI starts supporting other key types. Signed-off-by: Yann Dirson <yann.dirson@vates.tech>
- it is bad practice to "catch any" - not logging anything just hides information, which can be especially important here as the reason for a try/catch is not obvious (exceptions thrown by XenServerBackup.__init__?) Signed-off-by: Yann Dirson <yann.dirson@vates.fr>
Previous code structure was that of 2 nested if's, with nominal code path being in the positive branch of each if. Adding another condition will bring a need for logging the reason to ignore a backup, but doing that by converting inner `if` to an `elif` chain (arguably the simplest modification) would bring the nominal code path to switch to the negative/last branch of the inner `elif` chain. At the same time, the outer `if` would not seem special enough to deserve it special place (and the cyclomatic complexity). This commit leverages the existing `try:except:` block to switch to an "error out the loop" pattern, where the nominal code path is just linear. Using `StopIteration` may feel like an abuse, but: - it is the only standard `Exception` that is not a `StandardError` or a `Warning`, and defining a new one could seem overkill - the closest alternative would be a `while True` loop and breaking out in both exceptional code paths and at the end of nominal path, where the `break` statements would "stop the iteration" as well Signed-off-by: Yann Dirson <yann.dirson@vates.fr>
…rver#66) Signed-off-by: Yann Dirson <yann.dirson@vates.fr>
All files matchin `custom.*\.conf` will be kept upon upgrade. See: xapi-project/sm#600 Signed-off-by: BenjiReis <benjamin.reis@vates.fr>
The swraid code path does disassemble pre-existing swraid when we're going to create a new one, but this would not be done when the selected primary disk is part of a pre-existing raid. Signed-off-by: Yann Dirson <yann.dirson@vates.tech>
There are debug traces in the code, but no way to see them without modifying the code. They're not even intrusive enough to justify filtering them out by default. Signed-off-by: Yann Dirson <yann.dirson@vates.tech>
When an install with a local SR on a raid device pre-exists, selecting a member of that RAID probed that device directly, and sfdisk would report partitions to analyze, which the kernel filtered out because of the RAID. This caused failure to access those partitions through unavailable block devices. Since the goal here is to warn the user of possible data loss, scan the RAID device for SR instead. Signed-off-by: Yann Dirson <yann.dirson@vates.tech>
This is a first step before calling this code earlier. Signed-off-by: Yann Dirson <yann.dirson@vates.tech>
The former actually does some checks which can cause an upgrade to abort (although with very low probability), and running the latter first would leave a host in a broken state, in need for a Restore before attempting a new upgrade. Signed-off-by: Yann Dirson <yann.dirson@vates.tech>
This function searches for the /boot/efi mount point in the /etc/fstab file of the mount point directory passed as parameter and returns True if present, False otherwise. This will allow to check that an existing installation or backup boot mode matches with the installer boot mode. Signed-off-by: Thierry Escande <thierry.escande@vates.tech>
This patch checks the boot mode of an existing backups and disallows its restoration if it differs from the installer one. This uses the new helper function is_rootf_uefi(). Signed-off-by: Thierry Escande <thierry.escande@vates.tech>
In interactive mode, when Driver Disks are used as local media, they must *also* be installed as Supplemental Packs after installation. This change remembers the use of Driver Disks and makes sure the question is asked, even when the supplemental-packs feature is disabled. Signed-off-by: Yann Dirson <yann.dirson@vates.fr>
It may not be obvious to users that local-media Driver Disks must be inserted twice during installation. This change uses the memorization of usage of Driver Disks from previous commit, and uses it to explain users they likely wants to insert their Driver Disk again. Signed-off-by: Yann Dirson <yann.dirson@vates.fr>
While developing and/or installing from an already-verified ISO, yum's repo_gpgcheck just gets in the way. This provides a general kill-switch for checking repomd.xml.asc, through commandline or answerfile. Signed-off-by: Yann Dirson <yann.dirson@vates.fr>
Similar to no-repo-gpgcheck but for RPM sigs. Signed-off-by: Yann Dirson <yann.dirson@vates.fr>
Global *gpgcheck flag can change the default from True to False, and these new flags allow to override this default on a per-source basis. Signed-off-by: Yann Dirson <yann.dirson@vates.fr>
When originally implementing the per-source gpgcheck flags in answerfile[1], the full code was moved to an anwerfile-only location, breaking the original interactive-install implementation (and then following code review[2] the working code for interactive install disappeared further from the patch series). This moves the Repository flag-setting to `add_repos()` common code, while leaving the flag computation to the caller, since only the answerfile case has to do any non-trivial logic. - [1] 06b7007 - [2] #2 (comment) Signed-off-by: Yann Dirson <yann.dirson@vates.fr>
Signed-off-by: Samuel Verschelde <stormi@laposte.net> Orig-commit: 2731116
- show kernel-alt warning Originally-by: Rushikesh Jadhav <rushikesh7@gmail.com> - Update warning message when installing with kernel-alt - cleaner reboot action on that screen: return EXIT - fix detection of the kernel-alt boot parameter of the installer - install kernel-alt when booting on install-alt - create initrd and update grub configuration as post-install task Signed-off-by: Samuel Verschelde <stormi@laposte.net> Orig-commit: 243ce81 Orig-commit: d9a8a3e Orig-commit: 7d9e1b0 Orig-commit: 9447964
Signed-off-by: Yann Dirson <yann.dirson@vates.tech>
Originally-by: Samuel Verschelde <stormi@laposte.net> Signed-off-by: Yann Dirson <yann.dirson@vates.fr> Orig-commit: b20e303
The presence of this parameter changes the installer behavior, preventing local source installation and pre-populating the xcp-ng install url. Originally-by: Nicolas Raynaud <github@nraynaud.com> Use the `--netinstall` parameter that is automatically generated at preinit stage and translate it into `answers['netinstall'] = True`. Signed-off-by: Samuel Verschelde <stormi@laposte.net> Orig-commit: 36395c3 Orig-commit: 098a2f4 Orig-commit: 0b298cd
Signed-off-by: Samuel Verschelde <stormi-xcp@ylix.fr>
Signed-off-by: Yann Dirson <yann.dirson@vates.fr>
This complements the default choice previously changed in the TUI. Signed-off-by: Yann Dirson <yann.dirson@vates.fr>
…ntax.
This is a rewritten version of original RAID1 support found in early
8.3.0 versions, on top of upstream swraid support added ustream in
more recent installer versions.
It adds to upstream support:
- TUI support
- Answerfile support for the original XCP-ng syntax:
<raid device="md127">
<disk>sda</disk>
<disk>sdb</disk>
</raid>
Compared to original XCP-ng implementation, we lose:
- edge cases where the feature could be used to create a RAID for
other usages than the primary/system disk
- ability to put more than 2 disks in the RAID (though the TUI still
allows to select them)
Originally-by: Nicolas Raynaud <github@nraynaud.com>
Signed-off-by: Yann Dirson <yann.dirson@vates.tech>
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
4 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
This is a first iteration, featuring: