Prevent Docker from closing connection after 10 minutes with slow mongosh commands?

[theDanielJLewis]

I'm using Coolify to manage my server, installing MongoDB via the Docker image as usual.

When I am in the database via mongosh, any slow commands (like deleting millions of documents or compacting a collection) sort of times out after 10 minutes with the error MongoNetworkError: connection 2 to 127.0.0.1:27017 closed (I'm SSH'd into the server, and thus the MongoDB is accessible via localhost). I say "sort of," because it seems most of the commands continue running, but I never get the response because of that error. For example, a db.collection.deleteMany() should return a JSON object with the count of how many documents were deleted.

Nothing I do on the MongoDB side fixes the problem. But I haven't had this problem when I ran MongoDB on a different server without Docker.

So based on some stuff I see online, I'm starting to suspect Docker could be the problem.

What should I check or change if, indeed, Docker is the cause for this apparent 10-minute limit?

Thanks in advance!

[djsisson]

@theDanielJLewis where are you using ssh from?

i.e from your home pc to your server, then running docker exec manually or doing it via the coolify website ui??

[djsisson]

if you ran mongosh localhost:2017 then yes

if the connection drops you would see it in the tunnel terminal

docker and coolify doesnt do any networking

[theDanielJLewis]

Sorry, if there's no networking by Docker or Coolify (or the Coolify proxy), then what is exposing mongodb through port 5000?

Maybe I did your tunnel thing incorrectly. Here's what I did:

1. In one terminal, run ssh -L 2017:localhost:5000 root@serverip from my PC.
2. In another terminal, run mongosh "mongodb//user:pass@locallhost:2017/db" from my PC.
3. In the mongosh CLI, run my command.
4. Command loses connection after 10 minutes even though it continues running in the background (like a deleteMany()).
5. Nothing changes in that first terminal window.

[djsisson]

if the tunnel doesn't have any errors then the connection isn't being dropped

can you try add ?maxIdleTimeMS=1200000
to the end of your connection string

//thats twenty minutes

[djsisson]

you can also check

cat /proc/sys/net/ipv4/tcp_keepalive_time

but thats usually 2 hours i think

[theDanielJLewis]

proc/sys/net/ipv4/tcp_keepalive_time is 2 hours

I went crazy and added &connectTimeoutMS=0&maxIdleTimeMS=0&socketTimeoutMS=0&waitQueueTimeoutMS=0 to my connection query and the command still times out after exactly 10 minutes.

I'm about to give up.