Description
Currently the proxy container deployed by the automation uses the JSON logging driver, which is the Docker default.
However, the downside of this driver is that when a container is recreated, the logs of the previous container are lost. While this does not happen for normal container restarts, it does happen when a container's configuration is changed (as containers are immutable, so changing the configuration requires a new container to be created from scratch). This is inconvenient, especially when the configuration change is to change the logging level, and generally it is undesirable.
We considered switching to the journald logging driver, which should solve this problem and generally work well for our needs.
This issue is about finalising this decision and, if we are happy with this, changing the automation to use the journald logging driver for future deployments.
Note for future development: the current version of Docker supports writing to multiple logging destinations so if we wanted to we could also consider a centralised logging collection solution like Fluentd or similar.
┆Issue is synchronized with this Jira Improvement by Unito
┆Components: Automation
┆Priority: Major