This patch instructs gitlab's omnibus image to configure logging to only
retain 5 files for each services which are not larger than 10MiB or
older than a day (default rotation time).
This should help to prevent random crashes/becoming unhealthy, which
seems to originate from a filled up tmpfs for logging.
Reference:
https://docs.gitlab.com/13.11/omnibus/settings/logs.html#runit-logs
This patch adds a telegraf client to the gitlab role, that can be
activated by defining `gitlab_telegraf`. It allows to collect all
prometheus metric with telegraf and store them in influxDB.
This is an experimental feature.
Authored by: Sheogorath
Co-authored-by: Sheogorath <sheogorath@shivering-isles.com>
This patch adds two new variables to the gitlab playbook, making the
memory limits of the container configurable and this way easier to
adjust to one's needs.
This patch will provide a new config option `gitlab_log_tmpfs` to
configure the docker container to setup a tmpfs instead of the bindmount
for the log directory of GitLab.
This should solve all headache around logs clogging up the disk but
might just trades it against memory problems. So far the test-run looks
good, but time will tell if it is a sufficent solution.
