🪬 .tarO(log) 🪬
I've been given a layout for today's deployment
It doesn't promise success, but deployment is definitely needed
🪬 .tarO(log) 🪬
# [ $davids.sh ] · message #221
-
@ [ $davids.sh ] · # 1234
After the problems with the fucking rabbit in the last releases, I need help from any mystical forces.
-
@ [ $davids.sh ] · # 1235
Who else needs a reading, write what it's for in the comments
-
@ Arsen IT-K Arakelyan · # 1236
Have you already switched from RabbitMQ to something else?
-
@ [ $davids.sh ] · # 1238
It's not that simple.
I'm just rebuilding everything to reduce the dependency on RMQ.
But yeah, I'm not going to use it anymore in my life.
-
@ Arsen IT-K Arakelyan · # 1239
And how do you rebuild?
Any ideas on what to switch to? Or how to isolate/improve RMQ?
-
@ [ $davids.sh ] · # 1241
Create queues once and don't delete them
Change the maximum number of RPCs on queues to direct calls in the code (distributed monolith)
Migrate the rest of the RPCs to gRPC
Deploy RMQ on bare metal machines outside of K8s
Write a script for quick deployment of a new RMQ cluster, so that in case of problems, we don't have to struggle, but can simply switch to a new cluster
Transition from real-time events to cron jobs with aggregation of necessary states (we actually only need real-time in 1 place, where it already exists)
So far, there are no ideas other than Kafka / Redpanda, but I don't like them either. The only other thing we're looking at is EMQX, but that's just a hypothesis.