MongoDB based transport

[SebastianStehle]

Hi,

I am working on a MongoDB based transport: https://github.com/SebastianStehle/Rebus.MongoDb.Transport/blob/main/MongoTransport.cs

The idea is to have a fallback when my customers do not have anything like Azure Queue or so on and have a small installation.

I use find and modify to ensure that no other worker fetches the message. But what happens when the process crashed after the commit has been completed? Is there something like a timeout header after which the message could be moved back to "unhandled"? e.g. with a timer.

[mookid8000]

Actually I've made a message queue implementation for MongoDB as well: https://github.com/rebus-org/MongolianBarbecue

I just never made a Rebus transport for it. But maybe you can get inspiration from it? It uses find-and-modify too, and then it attaches a timestamp to the modified document, effectively locking it for a fixed duration. This way, if the process crashes, the message will pop up as visible again, once the timeout has elapsed.

[SebastianStehle]

This way, if the process crashes, the message will pop up as visible again, once the timeout has elapsed.

Yes, this was also my idea. But this timeout should be defined by the application and I have not seen something like this in Rebus. Ideally it is the same timeout that is used for a single message to be processed. For example Google PubSub also ha a acknowledge timeout: https://cloud.google.com/pubsub/docs/reference/rest/v1/projects.subscriptions/acknowledge

[mookid8000]

Rebus.MongoDb will have a new transport implementation again in v8