Asyncpg in Docker - why can it crash with a No Route To Host error if there is definitely a route?

Question

Python

There is a group of containers that is raised by a similar file:

version: "3"
services:
  ff_corecomp:
    restart: always
    build: ./corecomp
    depends_on:
      - ff_postgres
  ff_postgres:
    restart: always
    image: "postgres"
    environment:
      - POSTGRES_PASSWORD=pass

ff_corecomp has the following pooling process for asyncpg:

await asyncpg.create_pool(
        dsn="postgres://postgres:[email protected]_postgres:5432/")

Suddenly, the code started to crash on this line. Traceback:

spoiler

File "./utils/db.py", line 13, in init_pool
ff_corecomp_1 | db.pool = await asyncpg.create_pool(
ff_corecomp_1 | File "/usr/local/lib/python3.8/site-packages/asyncpg/pool.py", line 398, in _async__init__
ff_corecomp_1 | await self._initialize()
ff_corecomp_1 | File "/usr/local/lib/python3.8/site-packages/asyncpg/pool.py", line 426, in _initialize
ff_corecomp_1 | await first_ch.connect()
ff_corecomp_1 | File "/usr/local/lib/python3.8/site-packages/asyncpg/pool.py", line 125, in connect
ff_corecomp_1 | self._con = await self._pool._get_new_connection()
ff_corecomp_1 | File "/usr/local/lib/python3.8/site-packages/asyncpg/pool.py", line 468, in _get_new_connection
ff_corecomp_1 | con = await connection.connect(
ff_corecomp_1 | File "/usr/local/lib/python3.8/site-packages/asyncpg/connection.py", line 1668, in connect
ff_corecomp_1 | return await connect_utils._connect(
ff_corecomp_1 | File "/usr/local/lib/python3.8/site-packages/asyncpg/connect_utils.py", line 663, in _connect
ff_corecomp_1 | raise last_error
ff_corecomp_1 | File "/usr/local/lib/python3.8/site-packages/asyncpg/connect_utils.py", line 652, in _connect
ff_corecomp_1 | con = await _connect_addr(
ff_corecomp_1 | File "/usr/local/lib/python3.8/site-packages/asyncpg/connect_utils.py", line 621, in _connect_addr
ff_corecomp_1 | tr, pr = await asyncio.wait_for(
ff_corecomp_1 | File "/usr/local/lib/python3.8/asyncio/tasks.py", line 483, in wait_for
ff_corecomp_1 | return fut.result()
ff_corecomp_1 | File "uvloop/loop.pyx", line 1974, in create_connection
ff_corecomp_1 | File "uvloop/loop.pyx", line 1951, in uvloop.loop.Loop.create_connection
ff_corecomp_1 | OSError: [Errno 113] No route to host

No route to host gave me exactly two ideas: the network inside the docker is routing something incorrectly, or the postges container has not risen.

The first idea disappeared quickly: I opened a shell in the ff_corecomp container, and pings to ff_postgres calmly come from it. The second idea also turned out to be wrong: by the time the ff_corecomp container is launched, postgres already says that it is ready to accept connections. What else could be wrong here?

Answer 1

I didn't quite understand what solved the problem, but it's one of the two options from this question.
Or it was a problem in TUN, and then it helped

sysctl net.bridge.bridge-nf-call-iptables=0
sysctl net.bridge.bridge-nf-call-arptables=0
sysctl net.bridge.bridge-nf-call-ip6tables=0

Or the problem was in firewalld

# Allows container to container communication, the solution to the problem
firewall-cmd --zone=public --add-masquerade --permanent

# standard http & https stuff
firewall-cmd --zone=public --add-port=80/tcp --permanent
firewall-cmd --zone=public --add-port=443/tcp --permanent
# + any other port you may need

# reload the firewall
firewall-cmd --reload

In any case, I applied both solutions - everything worked.

Answer 2

You need to describe access from one container to another using links in docker-compose.yml. Now access is cut by the docker.