How to let podman logs appear as systemd service logs?

[shibumi]

Following problem:

We are running rootless containers with log-provider journald enabled. The logging works, but the log do not appear in the log of the systemd service.

This means when I check the logs of the service via journalctl -u <service_name> I see the logs of systemd starting/stopping the service, but I do not see any container logs.

Instead, the container logs are logged with the container name.

Our systemd file looks like this:

# /etc/systemd/system/haproxy.service
[Unit]
Description=HAProxy service
Wants=network-online.target
After=network-online.target
RequiresMountsFor=%t/containers

[Service]
User=ha_proxy
Group=ha_proxy
Environment=HOME=/home/ha_proxy
Type=oneshot
RemainAfterExit=true
Restart=on-failure
ExecStartPre=+/usr/local/bin/generate-haproxy-config.sh
ExecStart=/usr/bin/podman kube play --replace --service-container=true /etc/systemd/system/haproxy.yaml
ExecStop=/usr/bin/podman kube down --force /etc/systemd/system/haproxy.yaml
TimeoutStopSec=70

[Install]
WantedBy=default.target

When I do journalctl -u haproxy.service I see the ExecStartPre logs and the systemd logs, but I do not see the logs of the container.

Nov 04 15:15:29 HOSTNAME systemd[1]: haproxy.service: Deactivated successfully.
Nov 04 15:15:29 HOSTNAME systemd[1]: Started haproxy.service - HAProxy service>
Nov 06 09:28:50 HOSTNAME systemd[1]: Starting haproxy.service - HAProxy service>
Nov 06 09:28:50 HOSTNAME generate-haproxy-config.sh[952998]: Hostname found in ....

When I check journalctl -b and search for my logs I find them:

Nov 06 09:28:51 HOSTNAME haproxy-container[953135]: [WARNING]  (3) : Proxy server_instances ...
Nov 06 09:28:51 HOSTNAME haproxy-container[953135]: [WARNING]  (3) : Proxy traffic...
Nov 06 09:28:51 HOSTNAME haproxy-container[953135]: [WARNING]  (3) : Proxy healthcheck...
Nov 06 09:28:51 HOSTNAME haproxy-container[953135]: [WARNING]  (1) : Exiting Master process...
Nov 06 09:28:51 HOSTNAME haproxy-container[953135]: [NOTICE]   (1) : path to executable is /u>
Nov 06 09:28:51 HOSTNAME haproxy-container[953135]: [NOTICE]   (1) : haproxy version is 3.2.

How do I get these container logs into the same systemd service context?

[shibumi]

I think that our systemd Type=oneshot might have to do with this, but I do not know how to fix this.

Switching to Type=simple or Type=exec does not work. Switching to Type=forking would require having a PID file, but podman kube play does not support a pid file flag (like podman run does).

What other options do we have here?

[shibumi]

I tried creating a LogNamespace for the service, but this did not lead to any good results. Same issue. The systemd service + generate script logs land in the log namespace, but the actual container logs do not.

What works is searching for the container logs via journalctl CONTAINER_NAME=<container name>, but this requires that the operator knows the container name (that might be random).