ConfigMap rabbitmq-server-conf k8s

[FaraSys]

Hi everyone!
I have 3 cluster k8s v1.24
In two cluster I install cluster-operator via command:

1. k create ns rabbitmq-system
2. kubectl apply -f https://github.com/rabbitmq/cluster-operator/releases/latest/download/cluster-operator.yml (v. 1.14.0)
3. k apply -f rabbitmq.yaml
   And I change settings in configMap rabbitmq-server-conf. via command:
   k apply -f cm.yaml -n rabbitmq-system. It work fine.

On the third cluster today I install cluster-operator via command:

1. k create ns rabbitmq-system
2. kubectl apply -f https://github.com/rabbitmq/cluster-operator/releases/latest/download/cluster-operator.yml (v. 2.0)
3. k apply -f rabbitmq.yaml
   When i change configmap with my settings, configmap returns to default values.

Why is this happening?

my cm:
apiVersion: v1
kind: ConfigMap
metadata:
name: rabbitmq-plugins-conf
namespace: rabbitmq-system
data:
enabled_plugins: '[rabbitmq_peer_discovery_k8s,rabbitmq_prometheus,rabbitmq_management].'

---

apiVersion: v1
data:
advancedConfig: |
[
{rabbit, [
{consumer_timeout, undefined}
]}
].
operatorDefaults.conf: |
cluster_formation.peer_discovery_backend = rabbit_peer_discovery_k8s
cluster_formation.k8s.host = kubernetes.default
cluster_formation.k8s.address_type = hostname
cluster_partition_handling = pause_minority
queue_master_locator = min-masters
disk_free_limit.absolute = 2GB
cluster_formation.randomized_startup_delay_range.min = 0
cluster_formation.randomized_startup_delay_range.max = 60
cluster_name = rabbitmq
userDefinedConfiguration.conf: |
total_memory_available_override_value = 1258291200
cluster_partition_handling = pause_minority
vm_memory_high_watermark_paging_ratio = 0.99
disk_free_limit.relative = 1.0
collect_statistics_interval = 10000
consumer_timeout = 3600000
kind: ConfigMap
metadata:
name: rabbitmq-server-conf
namespace: rabbitmq-system

[Zerpet]

The Operator will try to reconcile all dependent objects (ConfigMap, StatefulSet, Service, etc.) from the information in RabbitmqCluster.spec. This means that any changes to a dependent object will be overriden, as you are observing.

To configure RabbitMQ, you can use RabbitmqCluster.spec.rabbitmq.additionalConfig (see this link) and RabbitmqCluster.spec.rabbitmq.advancedConfig. The Operator will include your configuration in a dedicated config file inside RabbitMQ Pods.

I'm surprised you don't observe the same behaviour in 1.14. I believe there has been no changes for quite a while. I suspect there might be an additional factor causing this e.g. paused reconcilliation via label, error in reconcile loop, to name a few, that may be causing different behaviour in your 1.14 cluster.

[FaraSys]

@Zerpet Thank you for you answer/
That's right. In 1.14.0, the configuration does not change after my changes, but in 2.0, my changes immediately return to their original values.

Honestly, I don't quite understand how to correctly specify these settings. I have downloaded cluster-operator.yaml, but how should I change it?
I changed like this but it doesn't work. Look my screen

[FaraSys]

After applying this yaml file, a standard configmap is created again

[mkuratczyk]

Have a look at some of the examples, especially custom-config.

[FaraSys]

@mkuratczyk Thank you. This solved my issue.