Automate splunk monitoring console app for indexer clustering

[kashok-splunk]

Automate Splunk monitoring console app for indexer clustering

• Add cluster master URL to MC configMap
• Add MC as a searchhead on CM to get indexer info
• Automate the "Apply Changes" button on the monitoring console app when IndexerCluster custom resource is applied
• For multi-site get cluster-info and append "SPLUNK_SITE" and "SPLUNK_MULTISITE_MASTER" env variables to MC

[sgontla]

I think we need an additional conditional check here, just to make sure that the current CRs namespace same as the ClusterMasterRef. If there is a mismatch, we should not proceed here... Thoughts?

Also, GetServiceFQDN is not an option here, and it should be the Master URL always?

[kashok-splunk]

How will namespace similar to ClusterMasterRef? We can actually use any valid service/URL name or pod name, we just need valid service where we can connect to

[sgontla]

Are we supporting the case of indexers in one namespace and CM in another namespace? If so, how are we dealing with the two MCs in different namespaces? Are we officially supporting the deployments across the namespaces?

[gaurav-splunk]

No,I dont think so we are supporting indexers in a different namespace than CM. Check validateIndexerClusterSpec.

func validateIndexerClusterSpec(cr *enterprisev1.IndexerCluster) error {
	// We cannot have 0 replicas in IndexerCluster spec, since this refers to number of indexers in an indexer cluster
	if cr.Spec.Replicas == 0 {
		cr.Spec.Replicas = 1
	}

	// Cannot leave clusterMasterRef field empty or else we cannot connect to CM
	if len(cr.Spec.ClusterMasterRef.Name) == 0 {
		return fmt.Errorf("IndexerCluster spec should refer to ClusterMaster via clusterMasterRef")
	}

	// Multisite / multipart clusters: can't reference a cluster master located in another namespace because of Service and Secret limitations
	if len(cr.Spec.ClusterMasterRef.Namespace) > 0 && cr.Spec.ClusterMasterRef.Namespace != cr.GetNamespace() {
		return fmt.Errorf("Multisite cluster does not support cluster master to be located in a different namespace")
	}
	return validateCommonSplunkSpec(&cr.Spec.CommonSplunkSpec)
}

[sgontla]

Same as the case of Licence Master. I may be missing something here, but please let me know if we don't need FQDN in a certain cases?
Wondering how this is different from a case when we are dealing with the REST calls..

[kashok-splunk]

For rest calls we are using fqdn service, For license master, we are not using it anywhere. This is just for MC to add LM as its peer

[kashok-splunk]

How will namespace similar to ClusterMasterRef? We can actually use any valid service/URL name or pod name, we just need valid service where we can connect to

[kashok-splunk]

Integration tests passing locally

[smohan-splunk]

Integration tests passing locally

Manual int-test run due to ongoing circleci limitations

[gaurav-splunk]

No,I dont think so we are supporting indexers in a different namespace than CM. Check validateIndexerClusterSpec.

func validateIndexerClusterSpec(cr *enterprisev1.IndexerCluster) error {
	// We cannot have 0 replicas in IndexerCluster spec, since this refers to number of indexers in an indexer cluster
	if cr.Spec.Replicas == 0 {
		cr.Spec.Replicas = 1
	}

	// Cannot leave clusterMasterRef field empty or else we cannot connect to CM
	if len(cr.Spec.ClusterMasterRef.Name) == 0 {
		return fmt.Errorf("IndexerCluster spec should refer to ClusterMaster via clusterMasterRef")
	}

	// Multisite / multipart clusters: can't reference a cluster master located in another namespace because of Service and Secret limitations
	if len(cr.Spec.ClusterMasterRef.Namespace) > 0 && cr.Spec.ClusterMasterRef.Namespace != cr.GetNamespace() {
		return fmt.Errorf("Multisite cluster does not support cluster master to be located in a different namespace")
	}
	return validateCommonSplunkSpec(&cr.Spec.CommonSplunkSpec)
}

[kashok-splunk]

Integration tests after the lastest commit