wyre-ai/msp-claude-plugins/cluster-operations

Nutanix Cluster Operations

Overview

Cluster-level work spans three namespaces. clustermgmt is the core: clusters, hosts/nodes, cluster profiles, and storage container associations. prism covers Prism Central's own surface — tasks, categories, cluster registration state. lifecycle exposes LCM (Lifecycle Manager) inventory: what software/firmware versions each cluster runs and what upgrades LCM recommends. Each is reached through its own executor after discovery.

Key Concepts

Common Workflows

Cluster health snapshot

1. listOperations(namespace="clustermgmt", search="clusters") →

list-clusters operation.

1. clustermgmt_execute to list clusters; capture name, extId,

version, node count.

1. For each cluster of interest, get it by extId for config and

health detail, and list its hosts for per-node state.

1. Fold in open alerts from the monitoring namespace (see the

monitoring-aiops skill) for a complete health picture.

Recent task audit on Prism Central

1. Discover the list-tasks operation in prism.
2. prism_execute with an OData filter, e.g.

_filter="status eq 'FAILED'" plus a time-range predicate, ordered by start time.

1. Report each task's extId, operation type, error messages, and

timing.

Upgrade posture (read-only)

1. listOperations(namespace="lifecycle", search="entities") to find

the LCM inventory operation; lifecycle_execute to list current component versions per cluster.

1. Discover and read LCM recommendations to see upgrade candidates and

blocking pre-check conditions.

1. Stop there: performing upgrades, pre-checks that mutate state, and

node imaging are non-GET and blocked by read-only mode. Deliver the findings as an upgrade plan for a human to execute.

Gotchas

• The lifecycle namespace is huge (~110 operations). Always narrow

listOperations with a search term; paging the whole namespace wastes turns.

• Storage containers appear in two namespaces. clustermgmt holds

cluster-side container associations; storage is the primary container/volume surface. Prefer storage for capacity questions.

• Cluster registration is a prism concern. A cluster missing from

clustermgmt lists may simply not be registered with this Prism Central — check registration state via prism before declaring it down.

Related Skills

• api-patterns — discovery workflow, OData, read-only mode
• storage — containers, volume groups, capacity
• monitoring-aiops — alerts and capacity forecasting