Add documentation for running mongodb in the cloud

[Mastercorp]

What exactly do you mean? Something like using sacred with https://www.mongodb.com/cloud/atlas ? Are these not enough? https://sacred.readthedocs.io/en/latest/observers.html#authentication

[SumNeuron]

@Mastercorp I think (but am not sure) that @JarnoRFB means a wholesome example including MongoDB setup? (e.g. if someone starts using MongoDB because of Sacred and have limited idea of how to set up a server and opts for a cloud solution)

[JarnoRFB]

@SumNeuron Yeah that was basically the idea. I talked to some people that said that they use the FileStorageObserver, because it is easier to collaborate on... However, it is easier if you have a remote server that everybody can write to, but people also said they do not want to administrate a mongoDB. So some pointers where you can easily setup a mongoDB in the cloud. I already have a PR for a setup with docker, so that should complement it. Basically just saying you can choose one of x,y,z providers and start your containers there.

[SumNeuron]

@JarnoRFB I am new to Sacred and not the most experienced with MongDB, but docker sounds like the right way to go (for local).

I have found an article using MLab as a cloud solution. Most free options only offer 0.5GB which is not a lot.

However, to my current (and perhaps wrong) understanding, using an observer saves basically everything (especially if you are adding your own artifacts) - which is sort of the purpose of Sacred. This means any free solutions will quickly be consumed in terms of storage (so a local version might be better...)

[JarnoRFB]

@SumNeuron yes if you are working on a project alone that is definitely easier and cheaper. But for larger projects with many people working on the same thing you want to have a remote server. I use scaleway which gives you 50GB for ~5EUR a month, which is quite ok in my opinion.

Feel free to check out https://github.com/IDSIA/sacred/pull/400/. Happy to hear if more documentation if needed on this.

[SumNeuron]

@JarnoRFB yeah, I would love it (plus I think it will greatly lower the barrier to entry to many others)

Basically something like:

• Set-Up
• install MongoDB
• install Docker
• create new directory with sub-structure:

project/
-- experiment.py
-- mongodb.config.json
-- __init__.py

# experiment.py

# ...

@ex.automain
def main(...):
    # example of saving dummy data and adding artifact

    # example of adding dummy metric
    pass

• configure MongoDB settings (e.g. where you want the local data to be stored)
• start docker container in detached mode ( link to details where to learn more)

[JarnoRFB]

Thanks for the ideas!