postvec

Appearance

Backup and restore (CLI) ​

See backup and restore for what pg_dump carries, refresh_models() and the queue checks. The CLI restores cluster configuration and proves the host.

After the dump is restored:

bash
sudo postvec setup --database app --embedded
sudo postvec doctor --database app --deep

Remote mode, point at the same endpoints as before the dump:

bash
sudo postvec setup --database app \
  --grpc 10.0.0.20:33333 \
  --http https://10.0.0.20:22222 \
  --switch-mode
sudo postvec doctor --database app --deep

Remote mode also restores the postvec-server side: every node must serve the same enabled models and the same converters as before the dump. A node that lacks one fails that request with MODEL_NOT_LOADED while its peers answer, so the failure looks intermittent. refresh_models() reloads postvec's cache from the node list, and fleet covers inventory parity.

Expected

doctor exits 0. Worker heartbeat advances. 99-postvec.conf matches the restored databases in postvec.database.

Install matching packages for that PostgreSQL major first: packages. A missing name in postvec.database makes the worker fail and respawn until the list is edited and PostgreSQL is restarted.

On a container host, postvec doctor finds no pg_lsclusters cluster: Docker.