splunk / splunk-operator

Splunk Operator for Kubernetes
Other
210 stars 115 forks source link

Splunk Operator: SearchHeadCluster CustomResourceDefinition - Please provide a way to apply different resource limits and affinity to the deployer instance #1307

Open gjanders opened 8 months ago

gjanders commented 8 months ago

Please select the type of request

Bug

Tell us more

Describe the request

Currently the CustomResourceDefinition for the kind: SearchHeadCluster does not provide a way to provide different configuration between the deployer and search head cluster members.

Expected behavior

For the indexing tier we have 2 unique CRD's, (kind: IndexerCluster and kind: ClusterManager). I'd like a solution like this for the Deployer and SearchHeadCluster, alternatively at least the ability to control the affinity settings and the request limits (cpu/memory)

Splunk setup on K8S N/A

Reproduction/Testing steps N/A

K8s environment N/A

Proposed changes(optional) Provide a unique CRD for the Deployer and SearchHeadCluster Or an alternative option to achieve this goal

K8s collector data(optional) N/A

Additional context(optional) I do notice: # 1048 # 459 # 1138

Are similar, although my request is specifically around resource limits and affinity. However any "fix" will (hopefully) resolve all the issues mentioned

gjanders commented 8 months ago

While this seems to file as a bug by default, it's more of an enhancement request