Small VPS tuning

SimpleDeploy itself is light (about 60 MB resident for 10-20 small apps). On a 1 vCPU / 1 GB VPS, the limits you hit are usually Docker, the apps, and the SQLite write cadence rather than SimpleDeploy. A few config choices keep things smooth.

Recommended starting specs

• 1 vCPU, 1 GB RAM, 25 GB disk: minimum for a couple of nginx-class apps.
• 2 vCPU, 2 GB RAM, 40 GB disk: comfortable for 10-20 small services.
• 4 vCPU, 4 GB RAM, 80 GB disk: room for stateful services (Postgres, Redis) plus apps.

Add 10 GB headroom on disk for image churn.

Trim metrics retention

metrics.tiers defaults are generous. On a tiny host, shorten the long tail:

metrics:
  tiers:
    - { name: raw, interval: 10s,  retention: 30m }
    - { name: 1m,  interval: 1m,   retention: 12h }
    - { name: 5m,  interval: 5m,   retention: 3d }
    - { name: 1h,  interval: 1h,   retention: 30d }
    - { name: 1d,  interval: 1d,   retention: 180d }

Reduces DB size and write volume meaningfully.

Cap the log buffer

log_buffer_size: 200

Default is 500. On a small host with no remote log shipping, 200 is enough to debug recent issues.

Tune rate limits

Defaults assume a beefier host. On 1 vCPU:

ratelimit:
  requests: 100
  window: 60s
  burst: 25

This blocks bursty scrapers that would otherwise pin CPU.

Prune Docker regularly

Old images and dangling layers eat disk fast. Use the dashboard’s Docker > Prune or a weekly cron:

docker image prune -a -f --filter "until=168h"

Vacuum the database

SQLite reclaims space on VACUUM. The system DB backup includes a compact mode that strips metrics and stats before vacuuming. Schedule it weekly.

Avoid heavy backups on the same host

pg_dump of a multi-GB Postgres on a 1 GB host will swap. Push backups straight to S3 to avoid local intermediate files.

Watch the right metrics

On small hosts: disk used, swap used, deploy duration. CPU and memory are visible in the dashboard but disk is the silent killer.