Closed Kami closed 8 years ago
I'm closing this since this would probably break the installer.
One of the potential solutions might be adding /etc/init.d/nginx restart
to the end of the bootstrap script (or similar)? Since installer doesn't use the bootstrap script so it will only affected install.sh runs.
Well, crap. I think this was meant to squash one :bug:, but introduce another.
If nginx restarts in the middle of a run, the logging stream in st2installer stops as the connection is killed. So, it's nerfed here and restarted in the installer.
Should probably make the service_restart
attribute a variable dependent on the state of the installer (https://github.com/StackStorm/st2workroom/blob/ef8f8fde061ebcff117edf8e5ec04e6015635a32/modules/profile/manifests/st2server.pp#L58)
Installer refreshes when there's no connection for 5 seconds, so it should remediate itself. If it doesn't happen, it's probably an installer bug.
@emedvedev Yeah, although the problem is that it generates a new cert so if we restarted nginx during installer run user would be prompted to accept self signed cert again so this won't work.
I noticed this bug while testing installer changes - it looked like installer branch / revision was correctly checked out but the installer still used incorrect revision.
I believe the issue is related to initial puppet run not restarting nginx, but I could be wrong.
The reason why I suspect this is because when I manually test the changes (manually cloned / checked out branch, restarted nginx and st2installer uwsgi process) it works fine.