potentially positioning it – in a few years, once the implementation matures – as a superior database for a wide range of use cases.
This gave me pause. My impression, having seen and heard less and less of MongoDB in recent years, is that it’s on its way out the door completely, and soon. That it’s already irrelevant.
Is the author correct? Where is usage thriving, or at least holding steady?
Big enterprises. They move slowly, and at the time the choice was made, mongo was the option people had heard of. Good luck getting them to migrate now.
These have been broken in MongoDB for a long time. The consensus algorithm mentioned is broken, and has been broken for awhile. They are working on fixing it now, AFAIK, with an actual Raft implementation. See: Ark Consensus.
The idea of having replica sets is great, sure. But many databases have this and get it right. Having it and getting it wrong does not mean you got it right.
In addition, the oplog uses GTID-like position identifiers that are identical across all nodes in a replica set. This allows clients to seamlessly follow writes across the replica set as a logical unit, without thinking too hard about the physical topology or which nodes are primary.
GTID’s are (or have been) brokein in MongoDB. They are a timestamp with second-resolution. Using timestamps for distributed events has been known to be broken for a long time in distributed systems.
I think the idea of the post is nice, reflecting on some unique and useful things it’s done. But I felt like a few of these are like congratulating the slow kid for coming in last in a race. On top of that, getting a few things right doesn’t mean your data isn’t at huge risk in MongoDB. Oh, it’s probably better now, but I think not messing up a few things doesn’t pay for the sins of MongoDB past.
This gave me pause. My impression, having seen and heard less and less of MongoDB in recent years, is that it’s on its way out the door completely, and soon. That it’s already irrelevant.
Is the author correct? Where is usage thriving, or at least holding steady?
Big enterprises. They move slowly, and at the time the choice was made, mongo was the option people had heard of. Good luck getting them to migrate now.
These have been broken in MongoDB for a long time. The consensus algorithm mentioned is broken, and has been broken for awhile. They are working on fixing it now, AFAIK, with an actual Raft implementation. See: Ark Consensus.
The idea of having replica sets is great, sure. But many databases have this and get it right. Having it and getting it wrong does not mean you got it right.
GTID’s are (or have been) brokein in MongoDB. They are a timestamp with second-resolution. Using timestamps for distributed events has been known to be broken for a long time in distributed systems.
I think the idea of the post is nice, reflecting on some unique and useful things it’s done. But I felt like a few of these are like congratulating the slow kid for coming in last in a race. On top of that, getting a few things right doesn’t mean your data isn’t at huge risk in MongoDB. Oh, it’s probably better now, but I think not messing up a few things doesn’t pay for the sins of MongoDB past.