Mine: khatru (Go) with a badger store and a bluge search index, on a VM with a 20 GB volume. The stack and why: https://nostras.app/node
What bit me was all disk:
- The disk filled up three times. Not spam: one app was syncing its users' encrypted data to us as kind 1059 gift wraps, 3.3 GB of a 4.6 GB store. Watch what's growing by kind, not just the total.
- Deleting events didn't give the space back. Badger only frees disk when compaction rewrites its tables, and LMDB under strfry doesn't shrink its file either. Find out how compaction works before you need it.
- Two of our stores (analytics, push subscriptions) were writing to the container's filesystem instead of the volume and got wiped on every redeploy. After setup, redeploy once and check that everything survived.
What bit me was all disk:
- The disk filled up three times. Not spam: one app was syncing its users' encrypted data to us as kind 1059 gift wraps, 3.3 GB of a 4.6 GB store. Watch what's growing by kind, not just the total.
- Deleting events didn't give the space back. Badger only frees disk when compaction rewrites its tables, and LMDB under strfry doesn't shrink its file either. Find out how compaction works before you need it.
- Two of our stores (analytics, push subscriptions) were writing to the container's filesystem instead of the volume and got wiped on every redeploy. After setup, redeploy once and check that everything survived.