Skip to main content

When something breaks

Always in this order:

sve doctor                          # first: says what is failing and why
docker compose ps                   # who is healthy and who is not
docker compose logs --tail 100 api

Two things that look like faults and are not​

curl https://your-domain gives "connection refused" from the server itself. That is the router's NAT: the machine cannot reach itself through its own public address. Check it like this:

curl -k --resolve your-domain:443:127.0.0.1 https://your-domain

502 for a few seconds after recreating a container. The gateway is holding the previous address. It clears on its own, or immediately with:

docker compose exec gateway nginx -s reload

An update left something broken​

sve update rolls back on its own if anything is unhealthy, so normally you are already back on the previous version. If not:

sve rollback

Rolling back restores the version, not the data: the backup taken before updating is still where sve update said it left it, and restoring it is a decision a person makes, because it throws away everything that has happened since.

The panel will not load but the channels play​

They are separate things: broadcasting does not depend on the panel. Look at sve logs api.

You cannot find the answer​

Save the output of sve doctor and docker compose ps and send them to your provider: those are the two things always asked for, and they save a whole round of questions.