Open JoshuaCWebDeveloper opened 4 years ago
Your documentation said in a few places that re-running the setup script should fix any configuration issues. So all I tried yesterday was re-running the setup script successfully a few times, but that did not resolve the problem.
Today, I decided to try following the instructions in https://github.com/spantaleev/matrix-docker-ansible-deploy/blob/master/docs/uninstalling.md to uninstall completely and install fresh. This resolved my issue; after successfully running the setup script, the postgresql data directory was populated and the start script also ran successfully.
This would seem to be a glitch in the installation process that apparently can't be solved by merely re-running setup but that can be solved by starting over fresh.
Any idea if something happened during the first installation that may have caused this? Or something special that you did that may have interfered with Postgres?
There was one thing about the first install that was abnormal.
The first time I ran the setup-all
task, it ran fine all the way through without any errors; however, it was run on a low memory gpc instance, and subsequently running the start
task resulted in the process hanging due to low memory on the system. I ended up killing the start
task, cleanly shutting down the instance, and upgrading it to a larger instance type. After booting it up again, running the start
task failed because of the failing postgres container. However, other containers with mounted volumes (for example the matrix-appservice-discord
container) started up fine without any complaints about missing data.
I did configure a handful of the documented add-ons for the first install, but I didn't attempt anything that was outside of the documentation.
Earlier today, I ended up having to upgrade the instance a second time, and had no problems switching it back on after the upgrade (no missing postgres data directory or anything else missing from the containers or the filesystem). Personally, I wouldn't expect turning the instance off to cause data corruption in a docker container running on it...
Due to an unrelated issue, I had to do another fresh install again later today, it went fine and this error did not happen again.
I'm guessing the Postgres container never did manage to properly start the first time around and it somehow corrupted its data directory.
(unless something more strange is going on, but we haven't had such complaints before).
There's usually no data loss for containers, even with abnormal server reboot, so it wouldn't be something I would worry about.
I had the same error today because I attempted to use a symlink for /matrix
.
After installing and attempting to start for the first time the ansible playbook with default postgres settings (no reference to "postgres" in the vars.yml file), the
matrix-postgres
container will fail to start with the following error:An examination of the container's filesystem confirms that the data directory is in fact empty:
I don't see any mention of PostgreSQL in the installation instructions and the
README.md
seems to imply that it is configured by default. The output of the installation showed nothing that looked like an error relating to postgres:Entire installation log is attached here: ansible-install-debug-001.log
This is my first time attempting to install this deployment on a server, so it is entirely possible I did something wrong. Any help or advice would be appreciated.