Deployments don't lock the obs-api package

[eduardoj]

As we can see from the deployment log, the zypper lock is removed

TASK [unlock obs-api] ************************************************************************************************************************************************************************
changed: [obs] 

But it's not restored back afterwards...

[eduardoj]

The lock is restored afterwards, at least in successful deployments. 🤔

Regarding the description of the issue:

As we can see from the deployment log, the zypper lock is removed

TASK [unlock obs-api] ************************************************************************************************************************************************************************
changed: [obs] 

But it's not restored back afterwards...

In this specific case the zypper lock was not restored because the condition was not met. It was skipped. We can see it in the logs, in the last TASK [include_tasks]:

TASK [include_role : unset_apache_maintenance_mode] ******************************************************************************************************************************************
skipping: [obs]

TASK [include_tasks] *************************************************************************************************************************************************************************
skipping: [obs]

TASK [include_tasks] *************************************************************************************************************************************************************************
skipping: [obs]

PLAY RECAP ***********************************************************************************************************************************************************************************
obs                        : ok=7    changed=3    unreachable=0    failed=0    skipped=8    rescued=0    ignored=0

[hennevogel]

Sounds like it's related to #67 ?

[eduardoj]

Sounds like it's related to #67 ?

Hm... #67 is simply not implemented. There is only one task which records to GitHub. And it always records a successful result. When the deployment fails, the task is simply not run. See:

ansible-obs/includes/github-deployment.yml

Lines 7 to 10 in 2317faf

	- name: update github deployment
	shell: "bin/dotenv bin/ogd succeed --ref={{obs_running_version.stdout}}"
	delegate_to: localhost
	when: obs_running_version is succeeded

I guess that the right approach to make sure that the deployments always lock the obs-api package is to always perform this task, even if the deployment has failed. In other words, to remove the conditions which prevent the lock from being performed.

[danidoni]

I guess that the right approach to make sure that the deployments always lock the obs-api package is to always perform this task, even if the deployment has failed. In other words, to remove the conditions which prevent the lock from being performed.

After re-reading the ansible-obs playbook and thinking about it, I believe @eduardoj is right. Locking zypper should happen always.