Excess service restarts during zero-downtime upgrade
While upgrading GitLab according to https://docs.gitlab.com/omnibus/update/README.html#multi-node--ha-deployment-with-geo, there are two sections like:
For all nodes that run Unicorn, Sidekiq, or the
geo-logcursor
daemonHot reload
unicorn
,sidekiq
andgeo-logcursor
services:sudo gitlab-ctl hup unicorn sudo gitlab-ctl restart sidekiq sudo gitlab-ctl restart geo-logcursor
However if you scroll up into the gitlab-ctl reconfigure
output, you will notice that it already restarted unicorn
, sidekiq
, and geo-logcursor
services. The reconfigure step follows the apt-get install gitlab-ee
step.
Two related problems
- We don't need
reconfigure
to restart unicorn during zero-downtime updates, we just want tohup
it - We don't need to restart these services in the instructions if reconfigure already did it
We need to check if this also affects non-Geo contexts (it seems likely.)
Edited by Michael Kozono