RabbitMQ talk

rabbitmq_outline.md

• RabbitMQ
  • A Young Lady's Primer on Distributed Queueing
    • Concepts
      • Publishers
      • Consumers
      • The queue
    • "Networks" or "Why god hates you, and everything you think you know is wrong"
      • Exactly Once Delivery
      • More-than Once Delivery
      • Less-than Once Delivery
  • "Operating RabbitMQ" or "Why the fuck am I being paged"
    • Clustering
      • performance concerns
      • Mirroring queues
      • nodes
        • disk nodes
        • ram nodes
    • Where do the messages live?
      • Where in the cluster
      • Where on the nodes
        • RAM
        • "Mnesia" or "the worst named datastore ever"
  • Further reading

• Can we convert a RAM node to a disk node and vice versa? How can we monitor the conversion process?
• Clustering patterns
• Why do Sensu messages keep getting stuck on RabbitMQ?

@richard2191 Can you give me some more info on messages getting stuck in RabbitMQ? i.e. is it specific messages getting stuck, or the "queue is completely backed up and not dequeueing" stuff we see?

When should I use 🐰 vs Kafka?

Does it matter whether I use STOMP, MQTT, or Emily POST? (rdrr)

1. 👍 on the title 😄
2. how can we capacity plan and is ratio of disk nodes:RAM nodes a place one has flexibility?
3. what IS the caveat of that ratio changing and skewing one way or the other....like..what use cases need what more?

How do you decide an ideal configuration of ram vs. disc nodes?
Is it ever bad to run "rabbitmqctl stop_app" on a node assuming it's not the only ram or disc node? What should we check after running stop_app or start_app?
Is there a way to verify a ram node or a disc node has everything in memory or written to disc?
How can we safely migrate the node that seems to own the cluster (i.e. the cluster_name node)?

I meant "queue is completely backed up and not dequeueing" that we see