The collector runs unprivileged. Docker's json-file logs are root-only,
and access to the Docker socket is root-equivalent. Use the journald driver
to make container logs readable through systemd-journal membership.
The collector never touches the Docker socket, never runs as root, and never
joins the docker group.
Add the logging driver to the service in your Compose configuration:
services:
api:
# Keep your existing image and other service settings.
logging:
driver: journaldApply the configuration as an operator with Docker access:
docker compose up -dThis recreates the service with the new driver. Plan for the interruption.
docker logs keeps working.
To make journald the default for new containers, add log-driver to
/etc/docker/daemon.json. Preserve any other settings in that file:
{
"log-driver": "journald"
}Restart Docker as root:
systemctl restart dockerThe default applies to new containers only. Recreate existing containers to change their logging driver.
Replace <name> with the container name. Check journal access as aken:
sudo -u aken journalctl CONTAINER_NAME=<name> -n 5Then check collection and redaction. --container accepts the service name
as it appears in the journal, including a short name followed by ., _, or
- in the full name. For example, the Swarm service stack_api can match
stack_api.1.abc123 and other tasks. An exact name takes precedence; otherwise,
all matching names become separate sources:
sudo aken collect --container stack_api --dry-runYou can also pass a 12- or 64-character hex container ID. If no name matches, collection stops before reading with this error:
no container named "NAME" in the journal; known names: a, b, c
The error lists up to 20 known names, sorted. Use a listed name or its service
prefix. To list all names as aken, run:
sudo -u aken journalctl --no-pager -q -F CONTAINER_NAMEIf the journal knows no containers, the error gives a setup hint:
no container logs in the journal; is the docker logging driver journald? See https://github.com/akenhq/aken/blob/main/docs/docker.md
Configure the journald driver as described above and recreate the containers.
If a selected journal source returns no lines, collect names the user running
the collector in its permissions hint. If journalctl writes a diagnostic, the
hint includes ; journalctl said: <line>. In a live session, a docker_logs
job with empty stdout and a stderr diagnostic returns status error, for
example journalctl: No journal files were found.. Both messages use the
first non-empty stderr line, up to 200 bytes.
See Collect logs for time windows and the review screen.
All containers share the docker.service journald bucket. The default is
10000 messages per 30 s. A busy container can cause messages from other
containers to be dropped too.
To increase the burst limit, create
/etc/systemd/system/docker.service.d/journal.conf as root, creating its
parent directory if needed:
[Service]
LogRateLimitIntervalSec=30
LogRateLimitBurst=100000Reload the configuration and restart Docker as root:
systemctl daemon-reload
systemctl restart dockerEnsure /var/log/journal exists or set Storage=persistent under [Journal]
in /etc/systemd/journald.conf. Without persistent storage, logs vanish at
reboot and --since windows can come back empty.
The collector does not read /var/lib/docker/containers or use the Docker
socket or a socket proxy. A read-only socket proxy route is planned for a
later release.