Introducing a Local Experience for Atlas, Atlas Search, and Atlas Vector Search with the Atlas CLI

Yaj_Vikani · 2023-12-13T17:51:22.629Z

Yes, upon researching few blog posts, I have came to the same conclusion but couldn’t find any solution to that. I would appreciate if you have any suggestions for plausible solutions.

cc: @Jakub_Lazinski

Jakub_Lazinski · 2023-12-18T12:54:02.350Z

Thanks, @Yaj_Vikani and @Jake_Turner, for bringing this up, and apologies for the delayed response. The local Atlas development experience was not included in the Public Preview, as noted in our documentation’s known limitations section.

However, lately some progress has been made on this anticipated feature. With the newly released Atlas CLI 1.14.0 version, we’re now offering support for running the local experience from a container. We are currently refining dedicated documentation, but for now, here’s a quick guide for running the Local Atlas experience directly from Docker or Docker Compose:

Docker

Fetch the latest mongodb/atlas docker image with docker pull mongodb/atlas:latest
Start the Docker image in bash mode using docker run -p 27777:27017 --privileged -it mongodb/atlas bash . More options here.
Set up a local deployment with atlas deployments setup --bindIpAll --username root --password root --type local --force
To connect to the deployment from the host (outside the container), use: mongosh --port 27777 --username root --password root

Docker-Compose

Install docker-compose with brew install docker-compose
Navigate to your project folder and create a docker-compose.yml file with the content provided below.

services:
  mongo:
    image: mongodb/atlas
    privileged: true
    command: |
      /bin/bash -c "atlas deployments setup --type local --port 27777 --bindIpAll --username root --password root --force && tail -f /dev/null"
    volumes:
      - /var/run/docker.sock:/var/run/docker.sock
    ports:
      - 27777:27777

Start docker-compose with docker-compose up
Connect to the deployment from the host (outside the container) using: mongosh --port 27777 --username root --password root

Please let me know how it works out for you!

Igor_Prokopenkov · 2023-12-18T15:30:11.096Z

Hello, first of all, thank you for announcement of this exciting feature! This certainly makes lives of the developers much easier and allows to create a reproducible environments.
Quote: " 1. Navigate to your project folder and create a docker-compose.yml file with the content provided above."
Where do we get the contents of docker-compose.yml from please?

Jake_Turner · 2023-12-18T17:06:58.000Z

Jakub_Lazinski:

docker run -p 27777:27017 --privileged -it mongodb/atlas bash

This is working for me to recreate above

version: “3.8”

services:
mongo1:
image: mongodb/atlas:latest
ports:
- 27017:27017
privileged: true
entrypoint: atlas deployments setup --bindIpAll --username root --password root --type local --force

@Igor_Prokopenkov

Jakub_Lazinski · 2023-12-19T14:33:43.737Z

Apologies, I’ve edited my previous message and included the exmaple compose file.

Igor_Prokopenkov · 2023-12-27T09:35:53.998Z

Thank you, it seem to work just fine on docker-ce linux!
On question is regards to the compose configuration, how can I make sure that DB data will be persisted on the host like in this example for a regular mongo:

    volumes: 
      - /data/db:/data/db

not sure is this 100% relevant for atlas version.

The thing is that it is creating at the moment a new cluster each time one runs docker compose, but it is obviously not desired behavior, the desired behavior is to create cluster once and reuse it upon subsequent runs.

Igor_Prokopenkov · 2024-01-02T08:27:44.661Z

Also at the moment there is an issue there - if one stops docker container and resumes it again, cluster after resuming shows IDLE state and not responsible anymore. Stopping cluster manually with atlas deployments stop before stopping the container and than resuming when container is resumed with atlas deployments start fixes the issue but this seem to be fragile and not really a proper solution either.

Jakub_Lazinski · 2024-01-10T08:45:33.402Z

Thanks @Igor_Prokopenkov for raising these great questions.
Could you shed a bit more light on how are you using the docker-compose method here? Is it a CI pipeline or a local dev environment or sth else?

In meantime I’ll see how we can map a volumen to the container’s data directory without breaking the solution.

Re “Also at the moment there is an issue there - if one stops docker container and resumes it again, cluster after resuming shows IDLE state and not responsible anymore” - can you provide steps you’re taking here to make sure we’re on the same page?
Also, what happens when you try running atlas deployments start right after the container is restarted?

David_Vincent · 2024-01-10T14:30:57.832Z

Hi @Jakub_Lazinski, I’ve had similar issues, and can share my details. I appreciate your engagement and trying to get this all to work in a local experience.

docker-compose. We want to use an Atlas Docker container both for local dev and a CI pipeline.

A. CI pipeline. The CI pipeline seems doable; the Docker container spins up, we can create a deployment using command, and based on what we’ve seen, we should be able to connect, populate data, run tests on functions/endpoints accessing/manipulating that data, etc. Running this all once, in isolation, seems to work ok.

  datastore:
    image: mongodb/atlas:v1.14.0
    hostname: mongo_atlas
    container_name: mongo_atlas
    privileged: true
    command: |
      /bin/bash -c "atlas deployments setup --type local --port 27017 --bindIpAll --username root --password root --mdbVersion 7.0 --force && tail -f /dev/null"
    ports:
      - 27017:27017
    volumes:
      - ./run/docker.sock:/var/run/docker.sock

The one concern we have about CI is that there doesn’t seem to be any caching of the MongoDB binaries downloaded in step 2 of creation of the cluster. Not being able to cache costs money in the form of additional pipeline minutes.

B. Local dev. Local dev is the real issue for us. We want to persist data, so that as devs start and stop containers as they need to, they don’t have to keep recreating the entire database (losing whatever objects they may have populated on their own by running a mongorestore on shared data or whatever).

Or, in other words, we want to be able to create a dev deployment once, and basically restart it and reconnect to it whenever we start the container, storing the data on a persisted volume. Ideally, we could do all of this via a single command: create the deployment if it doesn’t exist, start if it does.

Any input on how to do that would be wonderful. Here is what we have tried so far.

We have kinda been able to persist the data to a mapped volume, but have not been able to access the same deployment once a container is stopped. And although we can see the data on the mapped volume, we haven’t been able to access that data again via the deployment (more on that below).

We tried giving the deployment the name dev. We searched for the .wt files in the Docker container, which led us to adding this volume mapping:

    command: |
      /bin/bash -c "atlas deployments setup dev --type local --port 27017 --username root --password root --mdbVersion 7.0 --force && tail -f /dev/null"
    ports:
      - 27017:27017
    volumes:
      - ./run/docker.sock:/var/run/docker.sock
      - ./data/atlas:/var/lib/containers/storage/volumes/mongod-local-data-dev/_data

In my own attempts, that does allow me to create the deployment and persist the data. When first upping my docker-compose, I get:

mongo_atlas  | 3/3: Creating your deployment dev...
mongo_atlas  | Deployment created!
mongo_atlas  | 
mongo_atlas  | connection skipped
mongo_atlas  | 
mongo_atlas  | Connection string: mongodb://root:root@localhost:27017/?directConnection=true

… and I can connect using that connection string as well as see the data in my ./data/atlas folder.

If I then stop my container and re-up it using the same docker-compose, I have gotten a variety of responses.

Obviously, that docker-compose has a command of atlas deployments setup, which might well be inherently problematic to rerun.

In any event, I might be wrong, but there seems to be some relationship to the time elapsed since I stopped the container and the error that I get.

I seem to first get this if immediately re-upping:

mongo_atlas  | Error: "dev" deployment already exists and is currently in "running" state
mongo_atlas exited with code 1

If I try to re-up rapidly after that, I get this:

mongo_atlas  | 2/2: Creating your deployment dev...
mongo_atlas  | Error: exit status 125: Error: network name mdb-local-dev already used: network already exists

If I get that “network already exists” error, it seems to repeat forever.

However, sometimes I have gotten this instead (which I think has happened when I have waited longer to re-up):

mongo_atlas  | 3/3: Creating your deployment dev...
mongo_atlas  | Error: failed to connect to mongodb server: connection() error occurred during connection handshake: auth error: sasl conversation error: unable to authenticate using mechanism "SCRAM-SHA-1": (AuthenticationFailed) Authentication failed.
mongo_atlas exited with code 1

When I get that error, I am eventually able to restart the deployment by upping that same docker-compose, but it seems to re-initialize things with default data, losing whatever I tried to persist.

Again, that docker-compose has a command of atlas deployments setup. If I try changing the docker-compose command to this:

command: |
        /bin/bash -c "atlas deployments start && tail -f /dev/null"

… and then immediately re-up after stopping my container, I get this message:

mongo_atlas  | Error: currently there are no deployments

Basically, no matter what I have tried, once the container has been stopped, I have not been able to subsequently access the same deployment I created the first time I ran the docker-compose and keep my data.

If there are different commands that I should be running, I’d be grateful to learn what they are – and some further documentation for maintaining local dev environments might be helpful.

Running the container and deploying manually. I can’t speak for @Igor_Prokopenkov, who may know more and have done more, but here is what I have done:

Execute docker run -p 27017:27017 --privileged -it mongodb/atlas bash
Set up a deployment via atlas deployments setup dev --type local --port 27017 --bindIpAll --username root --password root --mdbVersion 7.0 --force or similar

Running atlas deployments lists shows me this:

sh-4.4# atlas deployments list
NAME   TYPE    MDB VER   STATE
dev    LOCAL   7.0.4     IDLE

And I can connect using the connection string.

If I then run atlas deployments pause dev before stopping the container, then atlas deployments list gives me this both before I stop the container and after I restart it:

sh-4.4# atlas deployments list
NAME   TYPE    MDB VER   STATE
dev    LOCAL   7.0.4     STOPPED

And after restarting, I can run atlas deployments start dev and reconnect.

However, if I simply stop the container without first running atlas deployments pause dev, then I get this when running atlas deployments list after re-upping:

sh-4.4# atlas deployments list
NAME   TYPE    MDB VER   STATE
dev    LOCAL   7.0.4     IDLE

… but I can’t connect using the connection string. I also can’t pause it:

sh-4.4# atlas deployments pause dev
Error: exit status 255: Error: OCI runtime error: runc: exec failed: cannot exec in a stopped container

Note that it still continues to show as IDLE when I run atlas deployments pause list.

And if I try to run atlas deployments start dev, it hangs.

So we could follow an OS-agnostic dev approach by using the Docker image to run a dev deployment in this way, but remembering to pause the deployment every time before stopping the container – or even being able to – doesn’t seem feasible.

Igor_Prokopenkov · 2024-01-11T12:26:29.356Z

This is a very comprehensive description. Nothing to add really, just want to share my docker compose automation script to cover the manual steps mentioned above:

#!/bin/bash -x

# Function to stop Atlas deployment gracefully
stop_atlas() {
    echo "Stopping Atlas deployment..."
    atlas deployments stop
}

# Function to start Atlas deployment
start_atlas() {
    echo "Starting Atlas deployment..."
    atlas deployments start
}

# Trap SIGTERM and call stop_atlas
trap 'stop_atlas' SIGTERM SIGINT

# Check if the deployment exists and its state
deployment_status=$(atlas deployments list | grep 'LOCAL')
if [[ -z "$deployment_status" ]]; then
    echo "No local deployment found. Setting up..."
    atlas deployments setup --bindIpAll --username root --password root --type local --force
else
    if [[ $deployment_status == *"STOPPED"* ]]; then
        start_atlas
    fi
fi

while true
do
   #sleep 1000 - Doesn't work with sleep. Not sure why.
   tail -f /dev/null & wait ${!}
done

Igor_Prokopenkov · 2024-01-12T16:01:01.621Z

Sorry, worth to share how to use this script as well:

  atlas:
    image: mongodb/atlas:latest
    ports:
      - 27777:27017
    privileged: true
    volumes:
      - ./docker-compose/atlas/entrypoint.sh:/entrypoint.sh
      - /var/run/docker.sock:/var/run/docker.sock
    entrypoint: ["/entrypoint.sh"]

Jakub_Lazinski · 2024-01-18T08:33:46.373Z

Thank you both! Really appreciate the thorough description of the steps you’re taking.

It seems that there’re a couple of issues here:

When using docker-compose data is not being persisted across runs (or actually ups)
Wehn using docker run… the deployment is corrupted if not paused before restarting the container
When using docker-compose MongoDB binaries are not being cached, requiring its download with every run

Would it ublock you (for the time being) if #2 was fixed? Trying to figure out the priorities here.

In meantime, I’ll regroup internally to see what’s the best course of action to provide a quick help for you.

Igor_Prokopenkov · 2024-01-18T08:52:39.989Z

Thank you for the prompt response.
In regards to numer 1. It is preserved, at least for me (under ubuntu 22.04), but the one can’t control where the data is stored as in community mongo image.
For our team number 2 is definitely a priority since if cluster breaks it will becomes unusable and has to be re-created which leads to data loss same as it would be in the 1st case.
Can’t confirm number 3 under my ubuntu environment either.

Jakub_Lazinski · 2024-01-18T08:58:09.074Z

Thanks Igor,

For #2: can you also share the details of how you’re stopping and starting the container after the initial “docker run… bash”?

David_Vincent · 2024-01-18T11:43:53.836Z

#2 is also the priority for us. We think #1 might be doable if #2 is solved.

Jakub_Lazinski · 2024-01-18T15:30:03.942Z

@David_Vincent bumping up the other question.
For #2: can you share the details of how you’re stopping and starting the container after the initial “docker run… bash”?

Steffen_Langer · 2024-01-22T10:47:52.930Z

I would like to add some input to this discussion because we’d like to persist our local Atlas development environment, too.

For #2: Some team members use Macs, some Windows with Docker on a Vagrant machine (provider virtualbox), so for the latter, the Docker service runs on a virtual Linux system.
The containers are usually only stopped when we restart our computers. That, however, happens regularly, e.g. to apply operating system updates.

Rishi_Jung · 2024-01-23T06:00:05.347Z

Relating to the current issues, it seems I’m unable to establish a connection to a Local Mongo Atlas Docker container from another Docker container within my local environment. Is such connection unfeasible?
I am seemingly able to connect to the local Atlas container utilizing ‘mongosh’, indicative that the container is indeed accepting connections. Despite this, a connection between my application in a separate container, and the Mongo Atlas container remains unsuccessful.

pymongo.errors.ServerSelectionTimeoutError: atlasdb:27778: [Errno 111] Connection refused, Timeout: 5.0s, Topology Description: <TopologyDescription id: *****, topology_type: Single, servers: [<ServerDescription ('atlasdb', 27778) server_type: Unknown, rtt: None, error=AutoReconnect('atlasdb:27778: [Errno 111] Connection refused')>]>

Thank you

Omri_Keret · 2024-01-25T16:13:29.745Z

Persistency is a major setback to our team as well. If anyone managed to solve it, please share

Brayden_Tidwell · 2024-02-08T03:56:41.413Z

We are also running into all of the issues described here. I did manage to find a workaround for #2 (container restarts). You can intercept the container kill signal in your entrypoint script and pause the deployment:

atlas-entrypoint

#!/usr/bin/env bash

DEPLOYMENT_INFO=$(atlas deployments list | grep 'my_deployment')

if [[ $DEPLOYMENT_INFO ]]; then
    # Restart a deployment
    atlas deployments start my_deployment
else
    # Create a new deployment
    atlas deployments setup my_deployment --type local --port 27778 --username root --password root --bindIpAll --skipSampleData --force
fi

# Pause the deployment whenever this container is shutdown to avoid corruption.
function graceful_shutdown() {
    atlas deployments pause my_deployment
}
trap 'graceful_shutdown' EXIT

sleep infinity &
wait $!

docker-compose.yml

...
mongodb_atlas:
  container_name: 'mongodb_atlas'
  image: 'mongodb/atlas:v1.14.2'
  ports:
    - '27778:27778'
  privileged: true
  entrypoint: '/home/scripts/atlas-entrypoint'
  volumes:
    - './scripts:/home/scripts'

Haven’t tested it extensively but it at least works when you run docker compose down.
Importantly, this does not work if you use tail -f /dev/null as suggested in the docs. I had to switch that to the sleep and wait commands shown above.