Hi,
time for my 6 months report on upgrade bugs. Fedora 43 Edition :D
As you may know, i run a bigger cluster of fedora VM servers and
baremetal installations. Due to the headless nature, distro-sync is the
prefered methode. Pls no discussions, the "normal" upgrade path is too
dangerous in case of those issues and i know all your arguments from the
last upgrade report ;)
This is just a report to give feedback to get new ideas how to handle stuff.
Upgrading those servers showed the following issues, some are caused by
special circumstances, i mention them anyway.
1. several services do not start
Root cause:
the replacement of (Example) /usr/sbin/sshd with a symlink to
../bin/sshd did not work.
NONE of those replacements worked.
Proxmox VM, no special kernels required, happend 1 out of 100 (we still
have to update more servers, expect more numbers )
2. Systemd 258 does work or does not work with the same LTS kernel on
different systems
Root cause: unkown - inconclusive informations
Symptoms:
- Units freezes of PID1 unit while!!!! upgrading the system is still in
progress.
- reverting to an even older kernel fixed it also.
XenServer, LTS Kernel 5.x . (no bugzillareport because of this special
snowflake)
Does not happen on Proxmox VM with newer Kernel.
Workaround: using systemd 257 from Fedora 42 for now.
3. missing Housekeeping
collision between config files for user and groups of bin / daemon a.o.
Happens on all systems with more or less sys users.
bugzilla.redhat.com/show_bug.cgi?id=2479538
4. GDM failing to start gnome-session
Happend on different hw and gpus
Workaround: replace with lightdm
https://bugzilla.redhat.com/show_bug.cgi?id=2481856
5. PLYMOUTH Service failes to mount sysroot in initramfs
Happend once so far.
Special note: the initramfs did not change for a while now. I can only
imagine that something recreated the initramfs without letting us admins
know.
XenServer, LTS Kernel 5.x . (no bugzillareport because of this special
snowflake)
Solution: disabling it in the kernelline
###################
### Absolut NOGO Bugs ###
###################
Does Fedora have a upfront notice about those breakages and is it a
subscribable resource?
The happy msg on this is: Fedora so commonly used, it got important to
be more careful ;)
one way to handle this would be to introduce a new package or an "old"
package that backs the breaking service up for a release cycle, like
it's done with postgresql servers.
6. POSTGRESQL server update from V16 to V18
Never ever make more than 1 version jump with postgresql servers. the
imgrationtool does not handle it at all:
F43vm # postgresql-upgrade --upgrade-ids
postgresql - Upgrade data from system PostgreSQL version (PostgreSQL 17)
No support for 16 => you have to revert the database, export databases,
upgrade server and reimport it.
I'm glad i new about that pgsql issue upfront and made sqldumps. All
hail MYSQL livemigrationfeature ;)
7. DOVECOT 2.3.x to 2.4.3 switch
Dovecot changed the config format in 2.4.x and is not capable of
migrating the old config correctly. The online tool is too dump for it.
manual recreation was required.
If the testservers did not find this, a major production fail would have
been the consequence.
best regards,
Marius Schwarz
--
_______________________________________________
devel mailing list -- [email protected]
To unsubscribe send an email to [email protected]
Fedora Code of Conduct:
https://docs.fedoraproject.org/en-US/project/code-of-conduct/
List Guidelines: https://fedoraproject.org/wiki/Mailing_list_guidelines
List Archives:
https://lists.fedoraproject.org/archives/list/[email protected]
Do not reply to spam, report it:
https://forge.fedoraproject.org/infra/tickets/issues/new
