Hi,
Fixes were added in the 2 tasks that were mentioned:
1. The command that we run to check for updates:
2. Making sure that cache is up to date is done only once and not for each
package:
Thanks,
Dana
On Mon, Oct 12, 2020 at 10:52 AM Yedidyah Bar David <didi(a)redhat.com> wrote:
On Sat, Oct 3, 2020 at 1:04 AM Erez Zarum <erezz(a)nanosek.com>
wrote:
>
> Seems the problem, atleast part of it (because still, it doesn't get to
the part of creating the imgbase layer) is related to the /tmp/yum_updates
file.
>
/usr/share/ovirt-engine/ansible-runner-service-project/project/roles/ovirt-host-check-upgrade/tasks/main.yml
> yum check-update -q | cut -d ' ' -f1 | sed '/^$/d' >>
/tmp/yum_updates
> For some reason it also lists those packages:
> gluster-ansible-cluster.src
> gluster-ansible-infra.src
> gluster-ansible-maintenance.src
> gluster-ansible-roles.src
>
> Which do not exists, those are source RPMs, so the ansible playbook for
the host upgrade fails.
>
> I did a quick workaround and added an "egrep -v src"
> yum check-update -q | egrep -v src | cut -d ' ' -f1 | sed '/^$/d'
>>
/tmp/yum_updates
>
> So now the ansible-playbook doesn't fail, it does says the upgrade was
successful the host reboots, but it doesn't get upgraded.
>
> Also, i noticed that the playbook does make sure that the dnf cache is
up to date (update_cache is set to true) when first checking the ovirt-host
package, but it also does this for every single package in the task after,
so there's no need for update_cache there.
This is tracked in:
https://bugzilla.redhat.com/show_bug.cgi?id=1880962
>
> As a workaround to upgrade to 4.4.2 i have to reinstall every host with
oVirt node 4.4.2 as it seems the upgrade process is broken.
Adding Dana.
Would you like to open a bug about this? Not sure about the exact flow
- I guess it's not strictly about having src packages but about having
packages that are not re-installable (meaning, you installed them not
from a repo, or removed the repo after installation, or they were
removed from their repo, etc.).
Thanks and best regards,
--
Didi