Skip to content

TrinoContainer.get_connection_url() returns the internal port, not the mapped host port #1111

Description

@rusackas

BODY:
TrinoContainer.get_connection_url() builds the URL as trino://{user}@{host}:{self.port}, and self.port is the fixed container-internal port set in __init__ (8080 by default). It's never resolved through get_exposed_port(). Since Docker maps that to a random host port, the URL this method returns can't actually connect to anything.

Repro (4.15.0):

from testcontainers.community.trino import TrinoContainer
from sqlalchemy import create_engine, text

with TrinoContainer() as trino:
    print(trino.port)                             # 8080
    print(trino.get_exposed_port(trino.port))      # the real mapped port, e.g. 58092
    print(trino.get_connection_url())              # trino://test@localhost:8080 - wrong

    engine = create_engine(trino.get_connection_url())
    with engine.connect() as conn:
        conn.execute(text("select 1"))             # ConnectionRefusedError / TrinoConnectionError

Building the URL manually with get_exposed_port() instead of self.port connects fine.

For comparison, CockroachDBContainer/CrateDBContainer don't have this problem since they route through _create_connection_url(), which does resolve the exposed port. Looks like an isolated miss in the Trino module rather than a pattern elsewhere.

Suggested fix, matching how the other modules do it:

def get_connection_url(self):
    return f"trino://{self.user}@{self.get_container_host_ip()}:{self.get_exposed_port(self.port)}"

Happy to open a PR with this if useful. Found it while wiring up a nightly testcontainers job for apache/superset's db engine spec tests.

Metadata

Metadata

Assignees

No one assigned

    Labels

    No labels
    No labels

    Type

    No type

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions