OSOpenStack Learning Academy
INTERVIEW PREPARATION

OpenStack Question & Answer Bank

All 50 questions from the supplied senior interview guide, grouped by module.

MODULE A · Q01–Q10

Core Architecture & Foundations

Q01. What is OpenStack?
OpenStack is an open-source cloud infrastructure platform for building and operating Infrastructure as a Service (IaaS) clouds. It manages pools of compute, storage and networking resources through APIs and modular services, allowing users to provision VMs, networks and volumes on demand. It is not a hypervisor; it coordinates technologies such as KVM, Linux networking and storage backends.
Q02. What are the main benefits of adopting OpenStack?
OpenStack offers infrastructure control, open APIs, reduced dependence on a single public-cloud provider and flexibility to integrate hardware and storage technologies. It supports multi-tenancy, automation, scale-out architecture and control over data location and security boundaries. It can use commodity hardware and be tailored to workloads. These benefits must be balanced against the skills, hardware, upgrades, monitoring and lifecycle management needed to operate a private cloud.
Q03. Explain the primary operational differences between OpenStack and AWS.
AWS is a public cloud operated by Amazon; customers consume managed services while AWS operates the infrastructure. OpenStack is software deployed and operated by an organization or provider on its own or hosted infrastructure. OpenStack operators manage hardware, capacity, networking, storage, upgrades and availability. AWS offers a broad portfolio of managed services and a consumption model. OpenStack offers more infrastructure ownership and architectural control but requires greater operational responsibility. They are different operating models, not exact equivalents.
Q04. What are the key modular components that form OpenStack's core architecture?
Common services include Keystone (identity/catalog), Nova (compute), Neutron (networking), Glance (images), Cinder (block storage) and Swift (object storage). Horizon provides a dashboard. Placement tracks resource inventories and allocations used by Nova. Heat provides orchestration, while Ceilometer may provide telemetry. Deployments may also include Manila for shared file systems and Ironic for bare-metal provisioning. The service set depends on requirements.
Q05. How do different components in OpenStack communicate with each other?
Services primarily communicate through REST APIs over HTTP/HTTPS, commonly exchanging JSON. A client authenticates with Keystone, receives a scoped token, discovers service URLs in the service catalog and presents the token to APIs. Internally, services may communicate asynchronously through RabbitMQ. Databases persist service state. REST APIs, message queues and databases have different roles.
Q06. What is the role of the message broker (RabbitMQ) in an OpenStack cluster?
RabbitMQ commonly provides the AMQP message bus for asynchronous remote procedure calls and notifications between service processes. It routes messages to queues so components can request work without synchronous direct calls. Acknowledgements and durable queues support reliable handling when configured. RabbitMQ does not store authoritative cloud resource state—that is held in service databases—and it is not the user-facing API endpoint.
Q07. What is Red Hat OpenStack Platform (RHOSP) and how does it differ from upstream OpenStack?
Upstream OpenStack is the community-developed open-source project and release stream. Red Hat OpenStack Platform is Red Hat's enterprise product and support offering, built from selected OpenStack components with tested integrations, packaging, lifecycle policies, documentation and vendor support. It may include vendor-specific deployment workflows and certified integrations. Capabilities and supported architectures vary by RHOSP release.
Q08. What is a "Tenant" (or Project) in OpenStack?
A project is a logical resource and ownership boundary. Users receive access through role assignments, and resources such as instances, networks, volumes and images can be associated with a project. Projects separate teams, customers, applications or environments in a multi-tenant cloud. Older terminology used tenant; modern Keystone and CLI terminology generally uses project. A project is not a separate physical infrastructure.
Q09. What is a "Role" in OpenStack?
A role is a named set of permissions used by policy rules to determine what an authenticated user may do. Keystone stores role definitions and assignments, such as assigning member to a user on a project. Assignments can be scoped to a project or domain; some deployments use system-scoped roles. Effective access also depends on service policy and token scope.
Q10. How can you view the list of roles in an OpenStack environment via the CLI?
Use `openstack role list` to display roles known to Keystone. Inspect one with `openstack role show <role-name-or-id>`. To view assignments, use `openstack role assignment list` with filters such as `--user`, `--project`, `--domain` or `--names`. Role definitions and assignments differ: listing roles does not show which users have each role.
MODULE B · Q11–Q18

Compute (Nova) & Virtualization

Q11. What is Nova and what is its primary function?
Nova is OpenStack's Compute service. It manages instance lifecycle: scheduling, building, starting, stopping, rebooting, resizing, migrating and deleting. Nova coordinates with Placement for allocation, Glance for images, Neutron for networking, Cinder for volumes and Keystone for identity. The VM runs on a hypervisor, commonly KVM/QEMU. Nova orchestrates; it is not the hypervisor.
Q12. What is a Hypervisor and which ones does OpenStack natively support?
A hypervisor creates and runs VMs while sharing physical CPU, memory, storage and networking. KVM/QEMU is a common production choice with Nova. Other drivers and integrations have included VMware vSphere, Xen-based technologies and Hyper-V, depending on release and driver support. Nova manages lifecycle and scheduling; the hypervisor runs the guest. Verify support for the exact release rather than assuming every historical driver remains supported.
Q13. Explain the concept and architectural purpose of "Cells" in OpenStack Nova.
Cells v2 divide a large Nova deployment into logical groups of compute hosts and associated services/databases. This supports horizontal scaling, distribution of database and service load, and isolation. A typical design has an API-level layer and one or more compute cells. The API layer routes requests and tracks instance mappings; cell services manage instances in their cell. Cells are not availability zones, host aggregates or projects.
Q14. What is the operational difference between pausing and suspending a VM instance?
Pausing freezes a running instance while its memory state remains resident on the compute host, so host memory remains consumed. It suits a short interruption. Suspending saves instance state to disk and stops active execution, allowing host memory to be reclaimed; resuming restores it. Exact behavior varies by backend. Neither is the same as a clean guest OS shutdown.
Q15. What are the CLI commands used to pause and unpause an instance?
Use `openstack server pause <server>` and `openstack server unpause <server>`. Inspect with `openstack server show <server>`. For suspension, use `openstack server suspend <server>` and `openstack server resume <server>`. Commands require suitable scope and permissions; verify resulting task and power state.
Q16. What is a "Flavor" in OpenStack?
A flavor is a predefined compute resource profile used when creating or resizing an instance. It typically specifies vCPU, RAM, root disk and ephemeral disk, with optional swap, metadata and extra specs. Extra specs can influence scheduling by requiring host capabilities or traits. A flavor describes requested resources; it does not guarantee capacity. Nova and Placement determine whether a suitable host is available.
Q17. What is the architectural role of nova-conductor?
Nova-conductor coordinates operations between Nova services and compute nodes and intermediates database access that nova-compute should not perform directly. This helps protect the database and centralize coordination. It participates in instance-management workflows. It does not run VMs and is not a hypervisor. Its precise responsibilities have evolved across releases.
Q18. What is the function of the nova-scheduler and how does it make placement decisions?
Nova-scheduler selects a suitable compute host. It evaluates resource requirements and constraints against capacity and host characteristics. Modern Nova uses Placement inventories, traits and allocations; the scheduler filters and weighs candidate hosts. Requirements can include vCPU, RAM, disk, availability zone, aggregate metadata and traits. Scheduling can fail if capacity is unavailable, constraints conflict or resource data is stale.
MODULE C · Q19–Q25

Storage (Cinder, Swift, Glance)

Q19. What types of storage are supported by OpenStack, and what are their use cases?
Common models include ephemeral storage tied to an instance lifecycle; Cinder persistent block volumes; Swift object storage; Manila shared file systems where deployed; and Glance-managed VM images. Backends may be local, distributed or external, such as Ceph. Block suits virtual disks and databases, object suits API-accessed unstructured data, file suits shared file access, and ephemeral suits temporary instance storage.
Q20. Explain Cinder and its role in an infrastructure stack.
Cinder is OpenStack's Block Storage service. It provides APIs to create, manage, attach, detach, snapshot and transfer persistent block volumes. A volume appears to a guest as a block device. Cinder drivers integrate with backends such as LVM, Ceph RBD, iSCSI, Fibre Channel and vendor arrays. Cinder manages volume lifecycle; the backend stores the data.
Q21. How does Swift structurally and operationally differ from Cinder?
Swift is object storage: objects reside in containers and are accessed through an HTTP object API. It suits unstructured data such as backups, media and archives. Cinder is block storage: it provisions virtual block devices attached to VMs and used like disks. Swift clients work with objects and metadata; Cinder consumers work with block devices and file systems. Their access patterns and operational characteristics differ.
Q22. What is Glance?
Glance is OpenStack's Image service. It lets users and services discover, register, upload, share and manage images used to create instances. An image contains operating-system or application disk content plus metadata such as format, visibility and architecture. Nova obtains image information and data through Glance when building an instance. Glance provides image cataloguing and access; it does not run VMs.
Q23. Where can VM images managed by Glance be stored (what backends are supported)?
Glance separates image metadata from image-data storage through configurable stores. Depending on release and configuration, image data may reside in a local file system, Swift, Ceph RBD or supported external integrations; some deployments use shared file systems or other supported stores. The available list is release- and packaging-dependent. Check the Glance configuration and documentation for the deployed version.
Q24. How can you transfer ownership of a Cinder volume from one project to another via CLI?
Use a volume transfer request. The current owner creates one, commonly with `openstack volume transfer request create <volume>`, receiving a transfer ID and authorization key. The receiving project accepts it with `openstack volume transfer request accept <transfer-id> <auth-key>`, subject to policy and volume-state restrictions. Check the installed CLI help for exact syntax. Treat the authorization key as sensitive.
Q25. What is the advantage of using Ceph as a unified storage backend in an OpenStack environment?
Ceph provides distributed block storage through RBD, object storage through its object gateway and file storage through CephFS. OpenStack services can integrate with these interfaces, potentially consolidating storage, scaling capacity, supporting replication and recovery, and reducing separate storage silos. Benefits depend on design, network capacity, failure domains, tuning and operational expertise. Ceph is not automatically simpler or faster in every environment.
MODULE D · Q26–Q32

Networking (Neutron)

Q26. What is Neutron?
Neutron is OpenStack's Networking service. It provides APIs for networks, subnets, ports, routers, security groups and floating IPs. It connects instances to tenant or provider networks and can provide routing, DHCP, address allocation and security through configured backends and agents. Neutron supports legacy agent-based deployments and OVN. It abstracts network provisioning while translating requests into underlying network configuration.
Q27. What are the common network virtualization types available to tenants in Neutron?
Common types include VXLAN, GRE and Geneve overlays, which encapsulate tenant traffic across an IP underlay. VLAN uses 802.1Q tags and requires the physical network to carry the VLANs. Flat provides untagged connectivity on a configured physical network. Availability depends on ML2 type/mechanism drivers, physical mappings and operator configuration. Tenants can create only types exposed by administrators.
Q28. What is the structural difference between a Provider Network and a Tenant Network?
A provider network is created by or controlled by the administrator and maps to a physical network or segment, such as VLAN or flat. It often connects instances to a data-center network. A tenant network is created by a project user within permitted options and is often an isolated overlay such as VXLAN or Geneve. Provider networks depend on physical configuration; tenant overlays provide logical isolation over shared infrastructure.
Q29. What is the purpose of a Security Group in Neutron and how does it operate?
A security group is a virtual firewall policy applied to Neutron ports. It contains ingress and egress rules for protocol, port range and remote IP prefix or security-group reference. Common configurations are stateful, allowing return traffic for permitted connections. Enforcement depends on backend such as Open vSwitch or OVN. Security groups operate at virtual ports and complement physical firewalls.
Q30. Explain the role and function of the Neutron L3 Agent.
In legacy agent-based Neutron, the L3 agent manages virtual routers and implements routing and NAT, often using Linux network namespaces and virtual interfaces. It routes between tenant and external networks and can handle floating-IP NAT. OVN deployments commonly implement logical routing and NAT through OVN components instead. Identify whether the deployment uses legacy L3 agents or OVN.
Q31. What are Floating IPs in OpenStack and how do they function under the hood?
A floating IP is an externally reachable address associated with an instance's Neutron port, usually through a router connected to an external network. In a typical legacy deployment, the router performs NAT between floating and fixed IPs. This enables external access without changing the guest's internal address. The address must come from an external pool, and routing, security rules and upstream reachability must be correct. OVN implementation differs.
Q32. What is Open Virtual Network (OVN) and why is it used in place of legacy Neutron agents?
OVN is a network virtualization system providing logical switching, routing, ACLs and NAT. Neutron's ML2/OVN driver translates Neutron resources into OVN logical configuration. OVN uses northbound and southbound databases and local Open vSwitch integration to distribute configuration. It can replace several legacy agent functions in supported designs. It is a distributed control- and data-plane architecture, not one agent.
MODULE E · Q33–Q40

Identity, Interface & Orchestration

Q33. What is Keystone and what are its basic functions?
Keystone is OpenStack's Identity service. It authenticates users and identities, issues tokens, manages users, groups, projects, domains, roles and assignments, and provides the service catalog. A client receives a token scoped to project, domain or system context and presents it to other APIs. Keystone establishes identity and authorization context; individual services enforce policy.
Q34. What is the OpenStack Service Catalog?
The service catalog is information returned by Keystone identifying available services and API endpoints. Entries commonly include service type, region and endpoint interface such as public, internal or admin. Clients discover API URLs through the catalog rather than hard-coding them. The catalog is endpoint discovery information; it does not prove service health or authorize every operation.
Q35. Explain the difference between Authentication and Authorization in the context of Keystone.
Authentication answers 'Who are you?' Keystone validates credentials and issues a token. Authorization answers 'What may you do?' Role assignments and service policy determine permitted actions within project, domain or system scope. Successful login does not grant administrative access automatically. A user may authenticate successfully but receive an authorization error if token scope or role is insufficient.
Q36. What is Horizon?
Horizon is OpenStack's web dashboard. It provides a graphical interface for operations such as launching instances, managing networks, viewing images, creating volumes and managing identity resources, subject to permissions and enabled dashboards. Horizon communicates with OpenStack APIs using the user's context. It is a user interface, not the control plane itself.
Q37. What is OpenStack Heat?
Heat is OpenStack's orchestration service. It automates creation and management of infrastructure resources using declarative templates. Users describe desired resources and relationships; Heat coordinates API operations to create, update or delete a stack. It can manage instances, networks, ports and volumes when plugins are available. It supports repeatability, dependencies, parameters and outputs.
Q38. What is a Heat Orchestration Template (HOT) and what format does it use?
A HOT template declaratively describes resources, parameters, mappings, conditions, outputs and dependencies for a Heat stack. It uses YAML and a defined structure including `heat_template_version`. It describes desired infrastructure rather than imperative shell steps. Heat parses it and orchestrates service API calls. Supported versions and resource types depend on the Heat release and plugins.
Q39. What is Ceilometer and what is its main use case?
Ceilometer is an OpenStack telemetry data collection service. It gathers measurements and events about cloud resources through configured agents, notifications and polling. Data can support usage reporting, rating, monitoring or integration with other telemetry systems. Ceilometer is not itself a complete monitoring dashboard or billing product. Meters and architecture depend on release and configuration.
Q40. What is OpenStack Ironic and how does it plug into the compute layer?
Ironic is OpenStack's Bare Metal service. It provisions and manages physical servers rather than VMs. Hardware drivers and management interfaces such as IPMI or Redfish support inspection, power control, provisioning and cleaning. Ironic integrates with Nova so physical machines can be requested as compute resources. It suits workloads needing direct hardware access or specialized performance.
MODULE F · Q41–Q50

Advanced Operations & Troubleshooting

Q41. How do you assign a project/tenant to a user using the modern OpenStack CLI?
Access is normally granted by assigning a role to a user on a project. A typical command is `openstack role add --user <user> --project <project> <role>`. Verify with `openstack role assignment list --user <user> --project <project> --names`. The user, project and role must exist, and the administrator needs permission. This is more precise than saying a project is simply assigned to a user.
Q42. What is the difference between Live Migration and Cold Migration of an instance?
Live migration moves a running instance between compute hosts with very low intended interruption. It generally transfers or shares memory and execution state and requires compatible hosts, networking and storage. Cold migration moves an instance that is not running, usually requiring it to be stopped. It causes more downtime but has fewer live-state constraints. Requirements depend on hypervisor, storage, CPU compatibility and policy.
Q43. How do you generate or upload SSH key pairs to inject into an instance via CLI?
Generate a key locally with `ssh-keygen`, protect the private key and upload only the public key. Register an existing public key with `openstack keypair create --public-key <public-key-file> <keypair-name>`, or generate through the client with `openstack keypair create <keypair-name>`. Select it at server creation with `--key-name <keypair-name>`. The image must support injection, commonly through cloud-init, and network rules must permit SSH.
Q44. What is the function of the cloud-init utility inside an OpenStack instance image?
cloud-init initializes a newly launched guest. It reads metadata and user data and can create users, configure SSH keys, set hostnames, write files, install packages and run first-boot commands. OpenStack can provide metadata through the metadata service or config-drive. cloud-init runs inside the guest; Nova makes metadata available but does not perform all guest configuration itself. Behavior depends on image and datasource configuration.
Q45. If an instance is stuck in an ERROR state, what is your initial troubleshooting workflow?
Inspect `openstack server show <server>` for status, fault, task/power state, host and attached resources. Review Nova API, scheduler, conductor and compute logs. Determine whether failure occurred during scheduling, image retrieval, networking, storage attachment, hypervisor creation or guest boot. Check quotas, Placement capacity, image availability, Neutron port/security rules, Cinder volume and compute-host health. Preserve evidence before reset or deletion; recover only after identifying the failing layer.
Q46. What tool is commonly used to deploy production-ready, containerized OpenStack environments?
Kolla-Ansible is a commonly used project for deploying OpenStack services in containers using Ansible. It provides container images and automation for configuring and operating a containerized control plane. Operators define inventory and configuration, then use Ansible workflows to deploy and manage it. Other approaches include OpenStack-Ansible and vendor platforms. Selection depends on release support, architecture, requirements and team skills.
Q47. Explain the term "Overcommit Ratio" in OpenStack and why it is used.
An overcommit ratio allows a cloud to offer more virtual resources than physical capacity, assuming workloads will not all use their maximum allocation simultaneously. CPU overcommit is commonly expressed as vCPU capacity relative to physical CPU capacity; memory and disk policies may also use ratios or reserved capacity. It improves utilization but increases contention risk. Monitor workloads and tune ratios; excessive overcommit can cause poor performance or scheduling failures.
Q48. What is the role of the Placement API in modern OpenStack environments?
Placement tracks resource-provider inventories, traits, aggregates and allocations. Nova uses it to determine availability and record resources allocated to instances. Inventories include resources such as VCPU, MEMORY_MB and DISK_GB; traits describe capabilities. Placement supports accurate selection and accounting. It is distinct from nova-scheduler, which makes decisions using Placement data and configured filters/weighers.
Q49. How do you check the health status of all OpenStack services via the CLI?
No single CLI command proves every component is healthy. Start with `openstack service list` for registration and `openstack endpoint list` for endpoints. Use `openstack compute service list` for compute and `openstack network agent list` where Neutron agents are deployed. Then test APIs, inspect processes/logs and check databases, message queues and storage. In OVN deployments inspect OVN/OVS components too. Registered/enabled does not necessarily mean end-to-end operational.
Q50. What is the structural difference between an OpenStack "Public Endpoint" and an "Internal Endpoint"?
Keystone's catalog can publish different endpoint interfaces. A public endpoint is intended for client access from outside the internal service network; an internal endpoint is intended for access from the cloud's internal network. An admin endpoint may be provided for administrative use. These are catalog labels and URLs, not firewall rules. DNS, routing, TLS, proxies and access controls determine actual reachability.