MongoDB reliability called into question?

[Bri]

This post is a little old, but I keep coming across articles similar to this, where the ability of MongoDB to reliably read and write is called into question (also, this is in contrast to a technology called RethinkDB where the team took the opposite approach of the Mongo team in terms of development, which is also an interesting read, specifically about the recent acquisition and why it happened)

As a light user of Mongo I haven't had any issues, but I also haven't heard any from this community specifically. Thoughts?

https://news.ycombinator.com/item?id=11857674

[julian]

We've got large companies using Mongo as well as many of our hosted clients.

So far I haven't seen any issues with MongoDB. Keep in mind however, that our use case is not complex.

We host single databases on non sharded setups. In terms of data size, we're nowhere near big data.

[scottalanmiller]

My guess is that this is people who don't understand their database and have built it incorrectly for their needs. MongoDB under some circumstances is not supposed to return all records. If you want that guarantee you either need to architect your database for it or use a database that guarantees it under all conditions. This is not a reliability issue with MongoDB unless these guys produce something a lot more substantial that rumours. MongoDB is pretty good, we've had it in production at some decent scale for three and a half years and I use it in numerous places, not only with NodeBB.

If you are looking for random threads of anonymous people posting "concerns" about things that they don't understand about any database, you can find it. Everyone tied to one product likes to have "concerns" about the products that they didn't choose or don't support (or sell.) I would take this with a grain of salt. If you look at the rest of that thread, it's a joke. People who clearly don't understand data theory and use cases or schemas or relational data just ranting because they hate the idea of "new" even when it is quite old (NoSQL pre-dates relational data, I've been a NoSQL engineer since 1989 and it was OLD then - we just didn't this cool retro name for it.)

With any database, it's worth knowing if there are issues. But MongoDB is a leader and absolutely perfect for this use case. The problem with the people in that thread is that they are looking for relational systems that are not sharded and are full synchronous across shards all of the time, choose something that they didn't understand that was the wrong tool for the job and instead of admitting that they were confused and goofed on their selection they rail against the product.

[Bri]

@scottalanmiller said in MongoDB reliability called into question?:

NoSQL pre-dates relational data, I've been a NoSQL engineer since 1989 and it was OLD then

That's an interesting tidbit, what was it called before NoSQL and what were the technologies?

Something else I've been looking at are graph databases like Neo4j and Arango, and now I'm kind of anticipating that you're going to tell me that it's in the same boat (i.e. old tech with a new name)

[scottalanmiller]

@Bri said in MongoDB reliability called into question?:

@scottalanmiller said in MongoDB reliability called into question?:

NoSQL pre-dates relational data, I've been a NoSQL engineer since 1989 and it was OLD then

That's an interesting tidbit, what was it called before NoSQL and what were the technologies?

Well it didn't have a specific name, if you wanted to refer to it as a category you might have called them non-relational databases, but the idea of making them a category didn't exist. NoSQL refers to such a broad set of things that it doesn't mean much, just that it isn't relational theory. Even today, talking about NoSQL is a bit silly since it includes so many unrelated things. NoSQL only exists as a reaction to so many people thinking that relational was the only thing that there was an just making crazy assumptions. It's kind of a movement of "reconsider how you want to store data."

There were still lots of different kinds of things... flat files, hierarchical, graph and such. I'm not sure, but document databases might be newer.

[scottalanmiller]

@Bri said in MongoDB reliability called into question?:

Something else I've been looking at are graph databases like Neo4j and Arango, and now I'm kind of anticipating that you're going to tell me that it's in the same boat (i.e. old tech with a new name)

Nope, old with still with the old name Graph databases are known to go back to at least 1969.

[Nataly Tenkova]

This post is deleted!