I run many different web apps off a single VPS, a Linode with Debian. Deployment is pushing to a git repo on the server, which has a hook that simply kills the server. Supervisord ensures apps are restarted automatically. Caddy proxies everything based on domain.
Litestream backs up the databases automatically to S3 compatible storage, and linode offers automatic snapshots of the VPS.
My deployment process is git push and that’s it. If I need to setup a new webapp, I just clone the repo, add to supervisor and caddy config, and I’m good to go.
If I need to do CI, I use GitHub, which just git pushes to the server to deploy.
A client of mine is setup using normally 7 Linode VMs..
Due to ahem.. "issues" related to a network change in the DC, that's going to balloon out to around 10 or 12 for a little while, then shrink back down.
- 2x load balancers (1 is active for public traffic; backup for private traffic; 1 is active for private traffic; backup for public traffic);
- 3x DB cluster nodes (currently at 5 due to aforementioned issues);
- 2x Web/App nodes
Currently planning to adapt this into 3 distinct sites with simpler 1x web/app + 1x db per-site to move the redundancy point out to specific DCs rather than to individual roles within a single DC.
12 comments
[ 3.9 ms ] story [ 25.3 ms ] threadProduction is Argo+Helm+Kubernetes+public cloud provider
Development is bash+docker compose+local computer or remote docker host
Litestream backs up the databases automatically to S3 compatible storage, and linode offers automatic snapshots of the VPS.
My deployment process is git push and that’s it. If I need to setup a new webapp, I just clone the repo, add to supervisor and caddy config, and I’m good to go.
If I need to do CI, I use GitHub, which just git pushes to the server to deploy.
Everything running over CloudFlareTunnel.
It's boring technology - PHP+MySQL+little bit Node.JS
Are the applications running on it CPU intensive?
Power usage on idle is around 70-80 Watt, on normal usage is around 100W, on peak can go to 200-250W.
Apps are not CPU intensive except one that uses disk in inefficient way. That's why i make a ram drive (tmpfs), put data there and then get out.
Due to ahem.. "issues" related to a network change in the DC, that's going to balloon out to around 10 or 12 for a little while, then shrink back down.
- 2x load balancers (1 is active for public traffic; backup for private traffic; 1 is active for private traffic; backup for public traffic);
- 3x DB cluster nodes (currently at 5 due to aforementioned issues);
- 2x Web/App nodes
Currently planning to adapt this into 3 distinct sites with simpler 1x web/app + 1x db per-site to move the redundancy point out to specific DCs rather than to individual roles within a single DC.