Timeout from Metabase in container – but not from shell in container

I have a client’s Metabase Enterprise running in a container env I manage for them. Connecting to databases in other compose envs on the same host works (through a common docker compose network each), but I cannot connect to databases on another host (so far: MS SQL and MariaDB); instead, I get a timeout after 10s.

Trying in a shell in the Metabase container, I can not just ping the other host, but connect to DB ports there with nc: E.g. nc -v reports the connection to MS SQL as open, I see the MariaDB “banner” after connection. The connection comes up instantly as well, imperceptible delay.

Any ideas what could make Metabase network differently than other programs in the same container? We’re using the official image. The connection runs over a wireguard interface on the host system, but I think that should be transparent to the container?

Forgot to mention: Metabase also can log to the same host that has the databases through UDP/GELF through the logstash-gelf driver that I mount into the container & load into metabase. That connection runs over the same physical link as the database connection, but outside the wireguard tunnel.

if your customer has an enterprise license please contact help at metabase dot com since your customer has paid/priority support

my 2c: please don't deploy a production environment with a docker compose. Compose was meant to do tests, not to deploy environments that need to be there 24x7x365

Thanks – was an unrelated networking confusion after all anyway; topic can be deleted.