Question for Kubernetes Rabbitmq Cluster -> unfair message incoming/outcoming

[채병훈]

I'm deploying rabbitmq to a kubernetes cluster.

Following the guide https://www.rabbitmq.com/kubernetes/operator/quickstart-operator.html.

I set it up according to the document above and deployed it according to the configuration below. There are 2 nodes.

And I'm looking at the grafana monitoring information while doing a load test, and only one node handles incoming messages and the other node handles outcoming messages. Is this right ? Shouldn't it be treated fairly?

apiVersion: rabbitmq.com/v1beta1

kind: RabbitmqCluster

metadata:

name: rabbitmq

namespace: infra

spec:

replicas: 2

resources:

requests:

cpu: 1000m

memory: 4Gi

limits:

cpu: 2000m

memory: 4Gi

rabbitmq:

additionalConfig: |

cluster_partition_handling = pause_minority

vm_memory_high_watermark_paging_ratio = 0.99

disk_free_limit.relative = 1.0

collect_statistics_interval = 10000

persistence:

storageClassName: gp2

storage: "200Gi"

affinity:

nodeAffinity:

requiredDuringSchedulingIgnoredDuringExecution:

nodeSelectorTerms:

- matchExpressions:

- key: app

operator: In

values:

- rabbitmq

[Michal Kuratczyk]

It depends on your client apps to a large extent and you haven't provided any information about that.

If you have 1 publishing connection and 1 consumer connection, connected to two different nodes, you'd get 1 node handling ingress

and the other one handling egress.

Note: 2-node clusters are a bad idea.

--
You received this message because you are subscribed to the Google Groups "rabbitmq-users" group.
To unsubscribe from this group and stop receiving emails from it, send an email to rabbitmq-user...@googlegroups.com.
To view this discussion on the web, visit https://groups.google.com/d/msgid/rabbitmq-users/473e9f71-d5f0-41a3-8763-6384b762025dn%40googlegroups.com.

--

Michał

RabbitMQ team