We are curently running our community forums on a dedicated 2Gb VM using the Discourse 'all in one' docker image which runs several services for Discourse in one container (maybe with docker in docker).
Maintenance is both convenient and complex: you need to stick with the provided scripts/utilties from discourse. You can even trigger updates from within the Discourse web interface. but with some upgrades things can break (like a postgresql index ) and you suddenly have to debug for 1-2 hours.
The Discourse instance is important for the community, but the 2Gb VM is using all available RAM.
we could move discourse to the swarm cluster and use our 3 node linode manage postgresql server. But it would mean we have to investigate if the provided setup can be migrated or moved to the cluster with separating out the postgresql service. Which will be quite a bit of work and ongoing maintenance.
This needs discussion first.
We are curently running our community forums on a dedicated 2Gb VM using the Discourse 'all in one' docker image which runs several services for Discourse in one container (maybe with docker in docker).
Maintenance is both convenient and complex: you need to stick with the provided scripts/utilties from discourse. You can even trigger updates from within the Discourse web interface. but with some upgrades things can break (like a postgresql index ) and you suddenly have to debug for 1-2 hours.
The Discourse instance is important for the community, but the 2Gb VM is using all available RAM.
we could move discourse to the swarm cluster and use our 3 node linode manage postgresql server. But it would mean we have to investigate if the provided setup can be migrated or moved to the cluster with separating out the postgresql service. Which will be quite a bit of work and ongoing maintenance.