action #174457
closedcoordination #169654: [epic] Create test scenarios for Agama
Use a post script to connect to the installed system with Agama for unattended installation with remote workers
0%
Description
Motivation¶
Use a post script to connect to the installed system with Agama for unattended installation with remote workers:
scripts: {
post: [
{
name: 'enable-ssh,
body: |||
#!/usr/bin/bash
systemctl enable sshd
|||,
chroot: true
}
]
}
Acceptance criteria¶
- AC1: Agama can boot using powerVM in OSD and is reflected in job group.
- AC2: Be in sync with YaST developers in case will create tooling for this before we implement this workaround.
- AC3: Focus on s390x, but if powervm is enabled and it is possible to check we could evaluate.
- AC4: (Nice to have) Additionally we could fix that we don't get logs by handling errors when the screen is not visible.
Additional information¶
Besides to have ssh open in the installed system, for security sle 16 comes without root access allowed,
one way to hack this is doing: echo 'PermitRootLogin yes' > /mnt/etc/ssh/sshd_config.d/root.conf
but probably is much more clean will be to uninstall openssh-server-config-rootlogin (but need to be tested)
When the screen is not visible in order to capture errors and finish the node process properly (meaning we will get logs properly)
we can use the approach we used in puppeteer: https://github.com/jknphy/e2e-agama-playwright/blob/main/pages/installation-finished-page.ts#L12-L22
Updated by JERiveraMoya 26 days ago
another question to resolve here, can we import a profile with a post section during interactive installation to solve the problem also there?
Updated by JERiveraMoya 25 days ago
Hi Leo, some doubts:
How do we know that the post script work to reconnect to the system? https://openqa.suse.de/tests/16256104#step/reconnect_mgmt_console/9
It is not clear to me from what I see in openQA. If that booting bug in s390x is still affecting there would make sense, but does it work the script in x86_64 ?
(actually the fix for that bug was accepted). Please track the fix, check that we can connect from x86_64 remotely manually (as the automation doesn't do that) and ensure that test suite where we import the profile are not broken now beacause of the post script, maybe it would even help. Actually I noticed that we don't need this for non-remote architectures but it is ok I guess.
Anyway I merged your PR, looks good, please setup the jobs groups with new file name extension.
Updated by lmanfredi 24 days ago · Edited
Changed in job group 607 the new file name extension: from default_sle_no_scc.json
to default_sle_no_scc.jsonnet
.
Also re-triggered the tests:
sles_default_unattended
sles_full_disk_encryption
sles_lvm
for ARCHs: aarch64
, ppc64le
and x86_64
.
Updated by JERiveraMoya 12 days ago
- Tags changed from qe-yam-dec-sprint-fy25 to qe-yam-jan-sprint-fy25
Updated by JERiveraMoya 12 days ago
- Status changed from In Progress to Resolved