IBM Turbonomic ARM v8.x Administrator - Professional cheat sheet

Core components: the platform (Turbonomic instance) hosts the API, UI, market analytics engine, and supply chain; targets are added as separate connections that report to it.

Deployment topologies: single-node (all services on one VM), multi-node/HA (separate roles for platform, database, and analytics), and Kubernetes/OpenShift operator-based installs.

Persistent state lives in the embedded or external database (e.g., external SQL Server or PostgreSQL for multi-node); sizing the DB is a prerequisite for HA.

The supply chain is a layered graph: supply entities (hosts, datastores, clusters, nodes, storage) feed demand entities (VMs, pods, containers, applications).

Installation prerequisites include NTP/time sync, DNS resolution both ways, supported OS version, and outbound/inbound port access between nodes and targets.

The market analytics engine is the decision core: it computes actions by resolving contention for resources among entities in the supply chain, not by static thresholds.

Upgrade order in multi-node: back up the database, upgrade platform services and analytics, then re-verify target connections and policies.

Component health is visible via the platform/system status page and API health endpoints; failed services should be checked before adding targets.

Licensing is edition-based (e.g., Standard vs. Premium) and capacity-based; the license file defines which editions and which entity types (VMs, containers, storage) are covered.

License usage is measured against the licensed capacity metric per entity type; exceeding licensed capacity triggers a compliance warning in the UI.

User roles are typically Administrator, Site Administrator (scoped), and read-only/observer; role assignment controls which actions a user can execute vs. only view.

Local users and LDAP/AD or SSO (SAML/OIDC) are the supported identity paths; group mapping assigns roles by directory group.

Global configuration includes time zone, currency, and default action mode (Recommend, Manual, Automatic) - action mode is the platform-wide default before policy overrides.

Audit logging records who executed which action and when; retention and export are configured at the platform level.

Notification settings (email/SMTP, webhooks) determine how action approvals and system alerts reach administrators.

Backup includes the database and configuration; restore must match the same version before applying further upgrades.

A target is a connection to a hypervisor, cloud, container platform, storage array, or app-monitoring system; each target type has required credentials and a minimum permission set.

Discovery is periodic (target-specific interval) and can be triggered manually; the first discovery populates the supply chain, subsequent ones refresh it.

Credential types: read-only monitoring credentials vs. action/automation credentials; actions require the higher-privilege credential to be validated.

Credential validation status is shown per target; a failed validation blocks discovery and actions for that target.

Targets can be grouped into target groups for bulk policy application and scoping.

Supply chain health: entities appear/disappear as targets are added or removed; orphaned entities indicate a disconnected or deleted target.

Stitching combines data from multiple targets (e.g., VM to datastore to storage array) so that shared resources are modeled once; missing stitching causes duplicate or unlinked entities.

Proxy/collector requirements apply when the platform cannot reach the target directly (e.g., cloud or on-prem networks).

Policy types include automation (action mode), placement, scaling, scheduling, and constraint (capacity/headroom) policies; policies are scoped to entity groups and inherited by children.

Action modes: Recommend (suggest only), Manual (execute after approval), Automatic (execute without approval); the effective mode is the most specific applicable policy.

Placement policies control where workloads may move (e.g., allowed/denied clusters, anti-affinity, pinning); they constrain the market engine's action candidates.

Scaling policies set min/max replicas or VM size bounds and whether scaling is enabled for a workload.

Scheduling policies define time windows during which automation is allowed or suppressed.

Constraints include utilization headroom, capacity buffer, and compliance constraints (e.g., required storage tier); they define what counts as acceptable after an action.

Policy precedence: more specific scope (entity-level) overrides group-level, which overrides global defaults.

Automation safety: a policy in Automatic mode bypasses approval, so validate in Recommend/Manual before switching.

Resource metrics: CPU (MHz), memory (KB), storage (IOPS/throughput/latency), network, and cost; utilization is measured against the entity's capacity, not the host's.

Actions categories: resize (up/down), move (VM migration), provision (add host/node), suspend, start, and scale (containers/pods).

The engine evaluates actions by benefit (improvement in utilization/performance) and risk (likelihood of contention after the action); actions are ranked by net benefit.

A resize action is generated when the entity's utilization is outside the acceptable range defined by headroom/constraint policies for a sustained period.

A move action is generated when the host is over-utilized and a destination can absorb the entity without violating placement or constraint policies.

Container optimization uses workload controllers and node/cluster capacity; scaling decisions consider both resource utilization and replica bounds.

Action approval workflow: pending actions appear in the UI; approving executes the action via the target's automation credential.

Optimization is continuous; after an action executes, the engine re-evaluates and may generate follow-up actions if conditions persist.

Dashboards show supply chain health, pending actions, and utilization trends; the entity view shows per-resource utilization and history.

Audit and action history record executed actions, approvers, and results; use these to trace why an action did or did not execute.

Common failure modes: credential validation failure, target unreachable (network/firewall), insufficient permissions for actions, and time skew.

If actions are not executing, check: action mode policy, approval status, credential permission, and target connectivity in that order.

If entities are missing from the supply chain, check target discovery status, stitching, and whether the target was removed.

Log bundles can be collected from the platform for support; include the timeframe of the issue.

Alerts/notifications can be routed to email or webhooks; tune thresholds to avoid noise.

Performance issues: verify database sizing and node resources; the analytics engine is CPU/memory intensive during large discoveries.

The REST API supports querying entities, actions, and policies, and executing actions; API tokens/credentials are managed per user.

Webhooks push notifications for actions and events to external systems (e.g., ticketing, chat).

App-monitoring integrations (APM) supply application-level data to correlate workload performance with infrastructure.

Cloud and container integrations (e.g., Kubernetes operators) allow Turbonomic to manage clusters and cloud resources natively.

Maintenance: regular database backups, version upgrades in supported order, and credential rotation.

Health checks after maintenance: verify target validation, supply chain population, and that policies/action modes are unchanged.

Extensibility: custom scripts or external automation can call the API to approve or trigger actions programmatically.

Version compatibility: check the support matrix for target versions (hypervisor, Kubernetes, cloud) before upgrading either side.

Core components: the platform (Turbonomic instance) hosts the API, UI, market analytics engine, and supply chain; targets are added as separate connections that report to it.

Deployment topologies: single-node (all services on one VM), multi-node/HA (separate roles for platform, database, and analytics), and Kubernetes/OpenShift operator-based installs.

Persistent state lives in the embedded or external database (e.g., external SQL Server or PostgreSQL for multi-node); sizing the DB is a prerequisite for HA.

The supply chain is a layered graph: supply entities (hosts, datastores, clusters, nodes, storage) feed demand entities (VMs, pods, containers, applications).

Installation prerequisites include NTP/time sync, DNS resolution both ways, supported OS version, and outbound/inbound port access between nodes and targets.

The market analytics engine is the decision core: it computes actions by resolving contention for resources among entities in the supply chain, not by static thresholds.

Upgrade order in multi-node: back up the database, upgrade platform services and analytics, then re-verify target connections and policies.

Component health is visible via the platform/system status page and API health endpoints; failed services should be checked before adding targets.

Licensing is edition-based (e.g., Standard vs. Premium) and capacity-based; the license file defines which editions and which entity types (VMs, containers, storage) are covered.

License usage is measured against the licensed capacity metric per entity type; exceeding licensed capacity triggers a compliance warning in the UI.

User roles are typically Administrator, Site Administrator (scoped), and read-only/observer; role assignment controls which actions a user can execute vs. only view.

Local users and LDAP/AD or SSO (SAML/OIDC) are the supported identity paths; group mapping assigns roles by directory group.

Global configuration includes time zone, currency, and default action mode (Recommend, Manual, Automatic) - action mode is the platform-wide default before policy overrides.

Audit logging records who executed which action and when; retention and export are configured at the platform level.

Notification settings (email/SMTP, webhooks) determine how action approvals and system alerts reach administrators.

Backup includes the database and configuration; restore must match the same version before applying further upgrades.

A target is a connection to a hypervisor, cloud, container platform, storage array, or app-monitoring system; each target type has required credentials and a minimum permission set.

Discovery is periodic (target-specific interval) and can be triggered manually; the first discovery populates the supply chain, subsequent ones refresh it.

Credential types: read-only monitoring credentials vs. action/automation credentials; actions require the higher-privilege credential to be validated.

Credential validation status is shown per target; a failed validation blocks discovery and actions for that target.

Targets can be grouped into target groups for bulk policy application and scoping.

Supply chain health: entities appear/disappear as targets are added or removed; orphaned entities indicate a disconnected or deleted target.

Stitching combines data from multiple targets (e.g., VM to datastore to storage array) so that shared resources are modeled once; missing stitching causes duplicate or unlinked entities.

Proxy/collector requirements apply when the platform cannot reach the target directly (e.g., cloud or on-prem networks).

Policy types include automation (action mode), placement, scaling, scheduling, and constraint (capacity/headroom) policies; policies are scoped to entity groups and inherited by children.

Action modes: Recommend (suggest only), Manual (execute after approval), Automatic (execute without approval); the effective mode is the most specific applicable policy.

Placement policies control where workloads may move (e.g., allowed/denied clusters, anti-affinity, pinning); they constrain the market engine's action candidates.

Scaling policies set min/max replicas or VM size bounds and whether scaling is enabled for a workload.

Scheduling policies define time windows during which automation is allowed or suppressed.

Constraints include utilization headroom, capacity buffer, and compliance constraints (e.g., required storage tier); they define what counts as acceptable after an action.

Policy precedence: more specific scope (entity-level) overrides group-level, which overrides global defaults.

Automation safety: a policy in Automatic mode bypasses approval, so validate in Recommend/Manual before switching.

Resource metrics: CPU (MHz), memory (KB), storage (IOPS/throughput/latency), network, and cost; utilization is measured against the entity's capacity, not the host's.

Actions categories: resize (up/down), move (VM migration), provision (add host/node), suspend, start, and scale (containers/pods).

The engine evaluates actions by benefit (improvement in utilization/performance) and risk (likelihood of contention after the action); actions are ranked by net benefit.

A resize action is generated when the entity's utilization is outside the acceptable range defined by headroom/constraint policies for a sustained period.

A move action is generated when the host is over-utilized and a destination can absorb the entity without violating placement or constraint policies.

Container optimization uses workload controllers and node/cluster capacity; scaling decisions consider both resource utilization and replica bounds.

Action approval workflow: pending actions appear in the UI; approving executes the action via the target's automation credential.

Optimization is continuous; after an action executes, the engine re-evaluates and may generate follow-up actions if conditions persist.

Dashboards show supply chain health, pending actions, and utilization trends; the entity view shows per-resource utilization and history.

Audit and action history record executed actions, approvers, and results; use these to trace why an action did or did not execute.

Common failure modes: credential validation failure, target unreachable (network/firewall), insufficient permissions for actions, and time skew.

If actions are not executing, check: action mode policy, approval status, credential permission, and target connectivity in that order.

If entities are missing from the supply chain, check target discovery status, stitching, and whether the target was removed.

Log bundles can be collected from the platform for support; include the timeframe of the issue.

Alerts/notifications can be routed to email or webhooks; tune thresholds to avoid noise.

Performance issues: verify database sizing and node resources; the analytics engine is CPU/memory intensive during large discoveries.

The REST API supports querying entities, actions, and policies, and executing actions; API tokens/credentials are managed per user.

Webhooks push notifications for actions and events to external systems (e.g., ticketing, chat).

App-monitoring integrations (APM) supply application-level data to correlate workload performance with infrastructure.

Cloud and container integrations (e.g., Kubernetes operators) allow Turbonomic to manage clusters and cloud resources natively.

Maintenance: regular database backups, version upgrades in supported order, and credential rotation.

Health checks after maintenance: verify target validation, supply chain population, and that policies/action modes are unchanged.

Extensibility: custom scripts or external automation can call the API to approve or trigger actions programmatically.

Version compatibility: check the support matrix for target versions (hypervisor, Kubernetes, cloud) before upgrading either side.