Task 2.4: Storage

• AThe container's writable layer

  Incorrect

  After a crash the kubelet restarts the container with a clean state, so files written there are lost.

• BA claim on a PersistentVolume with the Retain policy

  Incorrect

  That keeps data well beyond the Pod's life and needs storage provisioning, which is more than this cache needs.

• CA hostPath volume on the node

  Incorrect

  hostPath ties the Pod to a node's filesystem and outlives the Pod; it carries security risks and is not meant for scratch data.

• DAn emptyDir volume in the Pod spec

  Correct

  An emptyDir lives as long as the Pod. A container crash does not remove the Pod, so the data survives restarts, and it is deleted when the Pod leaves the node.

Scratch space shared for the Pod's lifetime is what emptyDir provides; persistent storage is for data that must outlive the Pod.

• AReadWriteOnce

  Incorrect

  ReadWriteOnce allows read-write mounting by a single node, so the second Pod on another node could not write.

• BReadOnlyMany

  Incorrect

  ReadOnlyMany allows many nodes but only for reading.

• CReadWriteMany

  Correct

  ReadWriteMany lets the volume be mounted read-write by many nodes at once.

• DReadWriteOncePod

  Incorrect

  ReadWriteOncePod restricts read-write access to a single Pod in the whole cluster.

Access modes describe how many nodes or Pods may mount a volume and whether they may write; shared writes across nodes need ReadWriteMany and a storage backend that supports it.

• AA larger pool of PersistentVolumes created ahead of time by hand

  Incorrect

  Pre-creating volumes is static provisioning; it shortens the wait but still needs an administrator to keep the pool filled.

• BA StorageClass with a provisioner that the claims name

  Correct

  Dynamic provisioning creates a volume on demand when a claim names a StorageClass whose provisioner can create it.

• CA ResourceQuota on requests.storage in each namespace

  Incorrect

  A quota caps how much storage a namespace may claim; it creates nothing.

• DAn emptyDir volume in each Pod that needs storage

  Incorrect

  emptyDir is scratch space that disappears with the Pod, not a provisioned persistent volume.

StorageClasses describe kinds of storage and the provisioner that creates them, which lets claims trigger volume creation on demand.

• AThe three replicas share one claim created from the template

  Incorrect

  The template is stamped out once per Pod; replicas do not share one claim.

• BScaling down from three replicas to two deletes the third replica's claim at once

  Incorrect

  The default retention policy is Retain, so claims stay when the set scales down unless the policy is changed.

• CThe claims are created in the kube-system namespace

  Incorrect

  Claims are created in the StatefulSet's own namespace, next to the Pods that use them.

• DEach replica gets its own PersistentVolumeClaim

  Correct

  For each volumeClaimTemplates entry, every Pod in the StatefulSet receives its own claim.

• EA replacement Pod for a replica mounts the claim that replica used before

  Correct

  When a Pod is rescheduled, its volume mounts use the claims tied to its identity, so the data comes back with it.

StatefulSets give each replica stable storage: one claim per Pod per template, reattached when the Pod comes back, and kept by default when the set shrinks or is deleted.

• AA claim must be in the same namespace as the Pod that uses it

  Correct

  Claims must exist in the same namespace as the Pod using the claim.

• BClaims are meant for StatefulSets, not for plain Pods

  Incorrect

  Any Pod can use a claim in its namespace.

• CThe claim needs the ReadWriteMany access mode first

  Incorrect

  Access modes do not lift the namespace rule.

• DPersistentVolumeClaims are cluster-scoped and cannot be referenced

  Incorrect

  Claims are namespaced; that is why the namespace matters.

Pods reach storage through claims in their own namespace.

• AOne file named configmap.yaml that holds both keys

  Incorrect

  The ConfigMap object is not written out as one YAML file; its keys are projected one by one.

• BEnvironment variables named after the two keys

  Incorrect

  Keys become environment variables through the container's environment settings in the Pod spec, not through a volume.

• CTwo files, app.properties and log.level

  Correct

  Each key in the ConfigMap's data becomes a file name under the mount path, with the value as the file's content.

• DAn empty directory until the keys are listed under items

  Incorrect

  items is optional and places chosen keys at specific paths; a plain mount without it projects every key in the ConfigMap as a file.

A mounted ConfigMap turns each key into a file, which suits applications that read configuration files. The optional items field places chosen keys at specific paths, and mounted keys are refreshed after the ConfigMap changes.

• AReleased volumes are deleted automatically within an hour

  Incorrect

  With Retain, nothing deletes the volume automatically.

• BIt still holds the old data and waits to be reclaimed by an admin

  Correct

  Released means the claim is gone but the resource is not yet reclaimed; it is not available for another claim.

• CThe new claim must use the old claim's name

  Incorrect

  Names do not reconnect a released volume.

• DThe volume must be expanded before it can bind again

  Incorrect

  Size is not the problem; its released state is.

Retained volumes move to Released and wait for manual cleanup before they can be used again.

• AAs two namespaces, one for each disk type

  Incorrect

  Namespaces group objects; they do not choose storage types.

• BAs two node labels on the storage nodes

  Incorrect

  Node labels describe nodes, not the type of disk a claim gets.

• CAs two StorageClasses, chosen by storageClassName

  Correct

  StorageClasses describe classes of storage, such as performance tiers, and claims pick one by name.

• DAs two access modes on the same class

  Incorrect

  Access modes describe how volumes are mounted, not their tier.

Administrators describe the storage they offer as StorageClasses, for example by quality of service.