Initial logs of an application container are lost

[pajachiet]

When an application container starts, its logs are not streamed to logmatic until the next tick of logmatic-docker.

This behavior loses up to 30s of logs. (30s = default logmatic-docker interval between ticks)
It may not be a problem for long running containers, but this loses most of the logs for short jobs.

One solution could be to push all available logs on running containers, as proposed in this issue. A way to do that could be to use containers.attach rather than container.logs, as tried in this commit.

Another solution could be to use the since parameter, as tried in this commit.

A hack we currently do to get all the logs is to sleep 30 seconds in the container, before starting the main command…