cert_file cannot be empty when using HTTPS

[sed-i]

Bug Description

CI occasionally fails because grafana is restarted with TLS config before the cert is there.

To Reproduce

Run CI, on a github runner, during peak hours.

Environment

Latest.

Relevant log output

unit-grafana-0: 15:21:55 ERROR unit.grafana/0.juju-log Could not restart grafana at this time: cannot perform the following tasks:
- Start service "grafana" (cannot start service: exited quickly with code 1)
----- Logs from task 0 -----
2023-10-06T15:21:55Z INFO Most recent service output:
    (...)
    logger=migrator t=2023-10-06T15:21:54.801375601Z level=info msg="Starting DB migrations"
    logger=migrator t=2023-10-06T15:21:54.820319486Z level=info msg="migrations completed" performed=0 skipped=451 duration=1.213231ms
    logger=plugin.finder t=2023-10-06T15:21:54.877879657Z level=warn msg="Skipping finding plugins as directory does not exist" path=/usr/share/grafana/plugins-bundled
    logger=plugin.finder t=2023-10-06T15:21:54.877950259Z level=warn msg="Skipping finding plugins as directory does not exist" path=/var/lib/grafana/plugins
    logger=secrets t=2023-10-06T15:21:54.878271167Z level=info msg="Envelope encryption state" enabled=true currentprovider=secretKey.v1
    logger=query_data t=2023-10-06T15:21:54.880673728Z level=info msg="Query Service initialization"
    logger=live.push_http t=2023-10-06T15:21:54.884229819Z level=info msg="Live Push Gateway initialization"
    logger=infra.usagestats.collector t=2023-10-06T15:21:55.045567343Z level=info msg="registering usage stat providers" usageStatsProvidersLen=2
    logger=provisioning.alerting t=2023-10-06T15:21:55.046046756Z level=error msg="can't read alerting provisioning files from directory" path=/etc/grafana/provisioning/alerting error="open /etc/grafana/provisioning/alerting: no such file or directory"
    logger=provisioning.alerting t=2023-10-06T15:21:55.046117157Z level=info msg="starting to provision alerting"
    logger=provisioning.alerting t=2023-10-06T15:21:55.046130558Z level=info msg="finished to provision alerting"
    logger=grafanaStorageLogger t=2023-10-06T15:21:55.049674948Z level=info msg="storage starting"
    logger=ngalert t=2023-10-06T15:21:55.050287864Z level=info msg="warming cache for startup"
    logger=ticker t=2023-10-06T15:21:55.050445768Z level=info msg=starting first_tick=2023-10-06T15:22:00Z
    logger=ngalert.multiorg.alertmanager t=2023-10-06T15:21:55.050629073Z level=info msg="starting MultiOrg Alertmanager"
    logger=server t=2023-10-06T15:21:55.051370792Z level=error msg="Stopped background service" service=*api.HTTPServer reason="cert_file cannot be empty when using HTTPS"
    logger=ticker t=2023-10-06T15:21:55.052723626Z level=info msg=stopped last_tick=2023-10-06T15:21:50Z
    logger=tracing t=2023-10-06T15:21:55.052913931Z level=info msg="Closing tracing"
    logger=server t=2023-10-06T15:21:55.136665472Z level=error msg="Server shutdown" error="*api.HTTPServer run error: cert_file cannot be empty when using HTTPS"
    *api.HTTPServer run error: cert_file cannot be empty when using HTTPS
2023-10-06T15:21:55Z ERROR cannot start service: exited quickly with code 1

Additional context

No response

[PietroPasotti]

@sed-i I'm looking into this issue, but you apparently already (thought you had) fixed it.

can you confirm that this occurs in main? if not I need a call to see what lead you to your solution because that's what I'd have tried as well.