I85PREP
⌘ K
LPIC-3 High Availability and Storage Clusters study materials
LPIC-3 High Availability and Storage Clusters study resources

LPIC-3 High Availability and Storage Clusters practice questions

One question at a time. Build your LPIC-3 High Availability and Storage Clusters understanding.

First 50 of 200 questions. Each one lists the choices, the correct answer, and an explanation. The checker at the top of this page saves progress in this browser when it is available.

Continue in the practice tool above to load more questions, 50 at a time.

Question 1High Availability Cluster Management

An administrator must prevent a Pacemaker cluster from entering split-brain, where both nodes could concurrently write to a shared LVM volume. Which cluster mechanism is designed to remove a node's access by force?

  1. A STONITH fencing device Correct answer.
  2. A colocation constraint
  3. A clone resource
  4. A quorum device for vote counting

Explanation

STONITH (Shoot The Other Node In The Head) uses a fencing device to power off or reset a node, guaranteeing that it can no longer access shared storage before recovery continues. Pacemaker treats fencing as mandatory when shared data is involved, and stonith-enabled defaults to true.

Question 2High Availability Cluster Management

Which Corosync service tracks cluster membership and maintains the quorum vote count used by Pacemaker?

  1. sam
  2. cpg
  3. votequorum Correct answer.
  4. cmap

Explanation

The votequorum service maintains the membership view and quorum votes, and it notifies Pacemaker when quorum is gained or lost. The other services handle group messaging, runtime configuration storage, and availability management rather than vote accounting.

Question 3High Availability Cluster Management

A resource is defined with the identifier ocf:heartbeat:IPaddr2. Which resource agent class does this identifier belong to?

  1. LSB
  2. service
  3. systemd
  4. OCF Correct answer.

Explanation

Pacemaker agent identifiers follow the form class:provider:type, so here the class is ocf, the provider is heartbeat, and the agent is IPaddr2. OCF agents support parameters, monitoring operations, and roles, unlike the simpler LSB and systemd classes.

Question 4High Availability Cluster Management

Given the command pcs constraint order promote ms_drbd then start fs_drbd with default settings, what behavior does Pacemaker enforce?

  1. ms_drbd may be promoted only on the node that currently holds a booth ticket.
  2. Promotion of ms_drbd completes before fs_drbd starts, and fs_drbd stops before ms_drbd is demoted. Correct answer.
  3. fs_drbd and ms_drbd must always be placed on the same cluster node.
  4. fs_drbd may start before ms_drbd is promoted, and both resources may stop in either order because ordering is optional.

Explanation

Ordering constraints default to kind=Mandatory and symmetrical=true, so the specified start ordering also implies the reverse ordering for stop actions. Promotion therefore precedes filesystem start, and the filesystem is stopped before the DRBD resource is demoted.

Question 5High Availability Cluster Management

An administrator runs pcs stonith create fence1 fence_ipmilan pcmk_host_list="nodeA nodeB". What does the pcmk_host_list parameter declare?

  1. The node that currently holds the fence device's lock
  2. The cluster nodes that this fence device is able to power off Correct answer.
  3. The nodes that are permitted to join the cluster
  4. The resource agents that may be fenced by this device

Explanation

pcmk_host_list maps a fencing device to the cluster nodes it can control; pcmk_host_map is used when each host needs a different address or plug identifier. Without this mapping, Pacemaker cannot select an appropriate device for a node that must be fenced.

Question 6High Availability Cluster Management

Which command removes the implicit location constraint and any ban created by pcs resource move web_vip node2, allowing the resource to run on any eligible node again?

  1. pcs resource cleanup web_vip
  2. pcs constraint delete web_vip
  3. pcs resource enable web_vip
  4. pcs resource clear web_vip Correct answer.

Explanation

pcs resource move adds a temporary location constraint with a high score, and pcs resource clear removes that constraint plus any ban so the scheduler can place the resource freely again. By contrast, pcs resource cleanup only resets failure counts and operation history for the resource.

Question 7High Availability Cluster Management

Which command initializes the metadata area on a shared block device so that it can be used as an SBD fencing device?

  1. sbd -d /dev/mapper/sbd dump
  2. sbd -d /dev/mapper/sbd create Correct answer.
  3. mkfs.sbd /dev/mapper/sbd
  4. pcs stonith create sbd fence_sbd devices=/dev/mapper/sbd

Explanation

The sbd create action writes the SBD header and message slots to the specified device, after which the device can be referenced in the cluster's SBD configuration. A watchdog timer must also be configured so a node self-fences when it loses access to the SBD device.

Question 8High Availability Cluster Management

Which Pacemaker configuration makes a DRBD resource capable of running in a Primary/Secondary role so that it can be promoted?

  1. Marking the resource as a plain clone with clone-max=2
  2. Declaring the resource promotable, for example with the pcs promotable keyword Correct answer.
  3. Setting the resource's target-role to Started
  4. Adding a colocation constraint with the filesystem resource

Explanation

A promotable resource, formerly called a master/slave resource, is a special clone that supports the Master role and can therefore be promoted. A plain clone runs identical instances without roles, so no promotion is possible.

Question 9High Availability Cluster Management

In a three-node Pacemaker cluster, two nodes fail and the remaining node has no quorum. With no-quorum-policy left at its default value, what happens to resources on the surviving node?

  1. The node stops all resources and waits until quorum is restored. Correct answer.
  2. The node promotes every promotable resource to compensate for the missing votes.
  3. The node ignores the loss of quorum and keeps all resources running.
  4. The node immediately reboots itself to force fencing of the failed peers.

Explanation

The default no-quorum-policy is stop, so Pacemaker stops resources on a node that lacks quorum because it cannot know whether other nodes are still active. This protects shared data; alternatives such as ignore, freeze, and suicide must be selected explicitly.

Question 10High Availability Cluster Management

Two independent Pacemaker clusters in separate data centers must ensure that a shared service is active at only one site. Which component issues a renewable ticket that arbitrates site ownership?

  1. pacemaker-remote
  2. SBD
  3. booth Correct answer.
  4. corosync-qdevice

Explanation

Booth is a ticket-based arbitrator that runs outside the two clusters; a site may run ticket-dependent resources only while it holds and renews the ticket. This provides cross-site arbitration without extending a single Corosync membership across the WAN.

Question 11High Availability Cluster Management

Which Pacemaker resource meta-attribute determines how strongly the cluster prefers to keep a running resource on its current node instead of moving it to a node with a better location score?

  1. migration-threshold
  2. priority
  3. stickiness Correct answer.
  4. target-role

Explanation

stickiness adds a score to the node where the resource is currently running, so a larger value makes the cluster reluctant to move it elsewhere. The default of 0 lets the resource move to any node with a better score, while a value of INFINITY effectively pins it in place.

Question 12High Availability Cluster Management

An administrator runs pcs cluster standby node1 on a two-node cluster in which every resource is allowed to run on either node. What is the immediate effect?

  1. node1 enters standby and its resources move to node2 while node1 keeps its vote Correct answer.
  2. node1 is fenced and rebooted by the cluster
  3. all resources are stopped on both nodes until node1 is returned to service
  4. Corosync removes node1 from membership, so node2 runs without quorum

Explanation

Standby marks the node as unavailable for resource allocation while leaving Corosync and Pacemaker running, so the node still counts for quorum. Pacemaker migrates the node's resources to another eligible node if one exists.

Question 13High Availability Cluster Management

A cluster contains the constraint pcs constraint colocation add web_vip with web_server INFINITY. What does this constraint enforce?

  1. web_vip must run on whichever node runs web_server Correct answer.
  2. web_vip is started before web_server on the same node, but may run alone
  3. web_vip is banned from any node that runs web_server
  4. web_vip is cloned on every node where web_server is active

Explanation

A colocation constraint with a score of INFINITY is mandatory: the dependent resource must be placed on the same node as the resource it is colocated with. If web_server is not active anywhere, web_vip cannot start either.

Question 14High Availability Cluster Management

A resource is configured with migration-threshold=3 and failure-timeout=120s. It fails on nodeA three times within 90 seconds, while nodeB is healthy and permitted by the location constraints. What does Pacemaker do next?

  1. it stops the resource on every node until an administrator runs pcs resource cleanup
  2. it retries nodeA a fourth time and then keeps the resource there permanently
  3. it moves the resource to nodeB and keeps it off nodeA until the failure records expire Correct answer.
  4. it fences nodeA and lets nodeB start all clone instances after the reboot

Explanation

When the number of failures reaches migration-threshold, Pacemaker treats the node as unsuitable and moves the resource to another eligible node. The fail-count on nodeA remains until failure-timeout elapses without further failures, after which the resource may return.

Question 15High Availability Cluster Management

In a four-node cluster, a ping resource is created and then converted with pcs resource clone ping clone-max=2. How many ping instances run when the cluster is healthy?

  1. one instance, because clones need clone-node-max set explicitly
  2. two instances per node, for eight in total
  3. two instances in total, placed on two different nodes Correct answer.
  4. four instances, one on every node in the cluster

Explanation

clone-max limits the total number of clone instances across the whole cluster, so a value of 2 produces two running copies. clone-node-max would instead limit how many instances a single node may host.

Question 16High Availability Cluster Management

A cluster contains pcs constraint order start db then start app kind=Optional. How does Pacemaker treat these two resources?

  1. app is blocked from starting unless db is already active somewhere in the cluster
  2. app can start without db, but starts after db when both are being started Correct answer.
  3. app is stopped whenever db stops and cannot restart until db is active again
  4. the constraint is ignored completely unless db and app belong to the same resource group

Explanation

An optional ordering constraint only sequences start actions when both resources are being started together; it does not prevent app from running while db is inactive. A mandatory (default) ordering would make db a hard prerequisite for starting app.

Question 17High Availability Cluster Management

After a fault on nodeB is repaired, a resource still will not start there because its recorded fail-count has reached the migration threshold. Which command clears that history so Pacemaker re-evaluates the resource immediately?

  1. pcs resource move web_server nodeB
  2. pcs resource disable web_server
  3. pcs cluster unstandby nodeB
  4. pcs resource cleanup web_server Correct answer.

Explanation

pcs resource cleanup resets the resource's operation history and fail-counts, after which Pacemaker schedules fresh start and monitor operations. It also clears a resource that is stuck in a failed state without relocating it.

Question 18High Availability Cluster Management

An administrator sets maintenance-mode=true on a running cluster. Which behaviour applies while the attribute remains set?

  1. every resource is stopped on every node until the attribute is cleared
  2. running resources are left alone and failures do not trigger any recovery action Correct answer.
  3. resources keep running but all monitoring is paused, so crm_mon reports them as unknown
  4. Corosync shuts down and the cluster loses quorum until an administrator intervenes

Explanation

In maintenance mode Pacemaker suspends its control actions: it will not start, stop, or move resources, and it ignores failures that are detected. This is useful for planned work that would otherwise cause the cluster to react.

Question 19High Availability Cluster Management

A node has fencing configured with an IPMI device at level 1 and a switched PDU device at level 2, both able to fence that node. How does Pacemaker use these devices when the node must be fenced?

  1. level 1 devices are tried first, and level 2 is used only if level 1 fails Correct answer.
  2. only level 2 is used, because level 1 is treated as a test-only tier
  3. the device whose pcmk_host_list matches the shortest hostname is selected
  4. both levels are always executed together so the node is fenced twice

Explanation

Fencing levels group devices into ordered tiers: Pacemaker attempts the devices in level 1 and falls back to level 2 only when the earlier level cannot fence the node. This provides redundancy when a primary fencing path is unavailable.

Question 20High Availability Cluster Management

A two-node Pacemaker cluster uses a quorum device, with corosync-qdevice on each node talking to corosync-qnetd. Node1 becomes unreachable while node2 stays healthy. What is the typical result for node2?

  1. node2 keeps quorum because the qdevice contributes an extra arbitration vote Correct answer.
  2. the qdevice fences node2 because it cannot reach node1 over the network
  3. node2 holds one of two votes and therefore loses quorum immediately
  4. the qdevice temporarily manages the resources until node1 rejoins the cluster

Explanation

A quorum device adds an external vote, so a two-node cluster effectively arbitrates with three votes and the surviving node can hold a majority. This avoids the classic two-node situation where either side alone lacks quorum.

Question 21High Availability Cluster Management

Which setting in corosync.conf allows a two-node cluster to keep quorum when one of the two nodes is lost?

  1. quorum_members: 2 in the nodelist section
  2. wait_for_all: 1 in the quorum section
  3. stonith-enabled: true in the Pacemaker properties
  4. two_node: 1 in the quorum section Correct answer.

Explanation

The votequorum two_node option changes quorum calculation for a cluster with exactly two nodes, so the surviving node retains quorum after its peer is lost. Without it, a two-node cluster can lose quorum permanently when only one node remains.

Question 22High Availability Cluster Management

An OCF resource agent action completes successfully. Which exit code must the agent return to Pacemaker?

  1. 1
  2. 7
  3. 8
  4. 0 Correct answer.

Explanation

OCF_SUCCESS is defined as exit code 0 and tells Pacemaker that the requested action succeeded. Other codes report conditions such as a generic error, an unimplemented action, or a resource that is not running.

Question 23High Availability Cluster Management

What is the effect of running `pcs resource disable web_server` on a running Pacemaker cluster?

  1. It marks web_server as unmanaged and leaves it running on its current node.
  2. It stops web_server and blocks it from starting until re-enabled. Correct answer.
  3. It removes web_server from the CIB and releases its resources.
  4. It moves web_server to the node with the lowest load.

Explanation

`pcs resource disable` sets the resource's target-role to Stopped, so Pacemaker stops the resource and does not start it again while it remains disabled. Re-enabling the resource restores the normal target role and allows placement to resume.

Question 24High Availability Cluster Management

A cluster contains `pcs constraint order start db then start app kind=Mandatory`. What happens to app when db is stopped?

  1. app is migrated to the node where db was running.
  2. app is stopped because Mandatory order requires db to be active. Correct answer.
  3. app is promoted to Primary and takes over db's role.
  4. app continues running but cannot be restarted until db is active.

Explanation

A Mandatory ordering constraint enforces both directions of the dependency: app cannot start before db, and app must stop when db stops. Weaker order kinds do not impose the same stop requirement.

Question 25High Availability Cluster Management

Given `pcs constraint colocation add vip with web_server -INFINITY`, which placement is enforced?

  1. vip must not run on a node where web_server is active. Correct answer.
  2. vip must run on the same node as web_server.
  3. vip and web_server must run on different clusters.
  4. vip prefers the node with the highest score for web_server.

Explanation

A colocation score of negative infinity creates an anti-colocation rule: Pacemaker will not place vip on a node where web_server is running. Positive infinity would require the two resources to run together.

Question 26High Availability Cluster Management

An administrator runs `pcs node maintenance node2`. Which behavior applies to resources currently running on node2?

  1. They are stopped immediately and banned from node2.
  2. They are converted to unmanaged resources cluster-wide.
  3. They keep running, but Pacemaker does not start or stop them. Correct answer.
  4. They are migrated to another node and node2 is fenced.

Explanation

Per-node maintenance mode makes Pacemaker stop managing resources on that node while leaving their current state intact. Existing resources keep running, but the cluster does not start, stop, or recover them until maintenance is cleared.

Question 27High Availability Cluster Management

An administrator runs `pcs resource move web_vip node2 --lifetime=30m`. No other location constraints exist. What happens when the 30-minute lifetime expires?

  1. The cluster fences node2 to force web_vip back to node1.
  2. web_vip is stopped because the move constraint has expired.
  3. web_vip is permanently banned from node2.
  4. The temporary constraint expires and normal placement rules apply. Correct answer.

Explanation

`pcs resource move` creates a temporary location rule, and `--lifetime` gives it an expiry time. When the rule expires, Pacemaker removes it and evaluates stickiness, location scores, and other constraints again.

Question 28High Availability Cluster Management

A Pacemaker cluster has `no-quorum-policy=freeze`. What does Pacemaker do when the cluster loses quorum?

  1. It stops all resources on all nodes immediately.
  2. It ignores quorum and starts resources on every node.
  3. It fences every node that does not hold the quorum device.
  4. It freezes resource management until quorum returns. Correct answer.

Explanation

The freeze policy suspends resource actions when quorum is lost, leaving running resources in place. Pacemaker resumes normal management only after quorum is restored.

Question 29High Availability Cluster Management

Which Corosync votequorum option lets a cluster recalculate expected votes as nodes leave, so remaining nodes can retain quorum within a configured window?

  1. two_node
  2. auto_tie_breaker
  3. last_man_standing Correct answer.
  4. wait_for_all

Explanation

last_man_standing, together with last_man_standing_window, allows votequorum to lower the expected vote count dynamically as nodes leave. This can keep a subset of a larger cluster quorate, but it must be designed carefully to avoid split-brain.

Question 30High Availability Cluster Management

A node in a three-node Pacemaker cluster with SBD configured loses quorum while `no-quorum-policy=suicide` is set. What is the intended result?

  1. The node removes itself from the Corosync nodelist permanently.
  2. The node self-fences through the configured stonith device and reboots. Correct answer.
  3. The node freezes all resource actions and waits for quorum to return.
  4. The node continues running resources because SBD provides quorum.

Explanation

The suicide policy makes a node that loses quorum fence itself, typically by rebooting through SBD and the watchdog. This prevents the isolated node from continuing to write to shared storage while another partition may form a new cluster.

Question 31High Availability Cluster Management

An administrator creates a Pacemaker resource group with `pcs resource group add webgrp web_vip web_server`. Assuming no other constraints exist, how does Pacemaker place and order these resources?

  1. It colocates them, starts web_vip first, and stops them in reverse order. Correct answer.
  2. It requires an additional colocation constraint before any group members can share a node.
  3. It starts web_server before web_vip because group members are ordered by resource type.
  4. It allows each group member to run on a different node and starts all members in parallel.

Explanation

A Pacemaker resource group adds implicit colocation and ordering: all members run on the same node, start in the order listed, and stop in reverse order. No separate colocation constraint is needed for the group itself.

Question 32High Availability Cluster Management

A resource has exhausted its migration threshold, and `pcs status` shows a failed action with a non-zero fail-count. Which command clears that failure history so Pacemaker can try the resource again?

  1. pcs resource restart <resource>
  2. pcs resource cleanup <resource> Correct answer.
  3. pcs resource enable <resource>
  4. pcs cluster start --all

Explanation

`pcs resource cleanup` clears the resource's operation history and fail-count, optionally limited to one node with `--node`. It does not change the resource's configuration; it makes the resource eligible for scheduling again.

Question 33High Availability Cluster Management

A fence device is configured with `pcmk_host_map="node1:port1;node2:port2"`. What does this attribute tell Pacemaker?

  1. How to map each Pacemaker node name to its fence device port. Correct answer.
  2. Which nodes are preferred for placing the fencing resource itself.
  3. Which resource agents are permitted to run on each cluster node.
  4. Which Corosync rings each node should use for cluster traffic.

Explanation

`pcmk_host_map` maps cluster node names to device-specific port identifiers so one fence device can control multiple nodes. The mapping does not define placement, resource agents, or Corosync ring addresses.

Question 34High Availability Cluster Management

A monitor operation for resource `db` is configured with `on-fail=fence`. The monitor action fails on node2 while the resource is running there. What is Pacemaker's immediate response?

  1. It fences node2 and recovers the resource on another eligible node. Correct answer.
  2. It restarts the resource on node2 and increments its recorded fail-count for that node.
  3. It ignores the failure until the next scheduled monitor interval expires.
  4. It stops the resource but leaves node2 available for other resources to run on.

Explanation

With `on-fail=fence`, a failed operation triggers fencing of the node where the failure occurred. After fencing, the cluster can start the resource on another node that satisfies its constraints.

Question 35High Availability Cluster Management

An administrator sets the cluster property `stonith-action=off`. A fencing operation is then required for a node whose fence device supports both reboot and power-off actions. What does Pacemaker request?

  1. The device disables STONITH for all nodes in the cluster.
  2. The device reboots the node and then powers it off.
  3. The device skips fencing and only marks the node standby.
  4. The device powers the node off rather than rebooting it. Correct answer.

Explanation

`stonith-action` selects the action Pacemaker asks a fence device to perform; its default is reboot. Setting it to off or poweroff makes Pacemaker request a power-off, assuming the device supports that action.

Question 36High Availability Cluster Management

An administrator wants to configure resource `db` as a promotable clone with `pcs resource create db ... promotable`. Which resource-agent capability is required for this configuration to work?

  1. The agent must support promote and demote actions and a promotable role. Correct answer.
  2. The agent must provide a fencing device class and a host map.
  3. The agent must run as a clone with clone-max set to exactly one.
  4. The agent must implement only start, stop, and monitor actions.

Explanation

A promotable clone requires an agent that can run in a promoted role and implements the promote and demote actions. Without those capabilities, Pacemaker cannot manage promotion and demotion of clone instances.

Question 37High Availability Cluster Management

In `corosync.conf`, which `totem` setting defines how long a node may hold the token before the cluster treats the token as lost?

  1. `join`
  2. `token` Correct answer.
  3. `consensus`
  4. `retransmits`

Explanation

The `token` timeout is the maximum time the token may be held before Corosync considers it lost and begins failure handling. `consensus` and `join` control different membership timing, while `retransmits` controls retransmission count.

Question 38High Availability Cluster Management

Which Pacemaker cluster property controls how often the scheduler re-evaluates time-based rules and constraints even when no cluster event occurs?

  1. The `pe-input-series-max` scheduler setting
  2. The `transition-delay` scheduling setting
  3. The `cluster-recheck-interval` property Correct answer.
  4. The `stonith-timeout` fencing setting

Explanation

`cluster-recheck-interval` sets how often Pacemaker rechecks the CIB and re-evaluates rules, including time and date constraints, without waiting for a cluster event. Shorter values increase responsiveness at the cost of more scheduler activity.

Question 39High Availability Cluster Management

A monitor operation for resource `web` uses `on-fail=block`. The monitor fails on node1, and no other operation has failed. What does Pacemaker do with `web` on node1?

  1. It fences node1 immediately and lets the cluster recover web on another eligible node.
  2. It restarts web on node1 after the monitor interval expires.
  3. It marks web failed on node1 and blocks further web operations there. Correct answer.
  4. It ignores the monitor failure and leaves web running on node1 until an administrator intervenes.

Explanation

With `on-fail=block`, Pacemaker records the failure and prohibits further operations for that resource on the affected node, rather than fencing the node. Recovery elsewhere still depends on the resource's constraints and the cluster's scheduling decision.

Question 40High Availability Cluster Management

A node has been powered off manually after a failure, and Pacemaker is waiting for fencing confirmation before recovering resources. Which command tells Pacemaker that the manual fencing of node2 is complete?

  1. `pcs node unstandby node2`
  2. `pcs cluster start node2`
  3. `pcs stonith confirm node2` Correct answer.
  4. `pcs resource cleanup --node node2`

Explanation

`pcs stonith confirm node2` acknowledges that the named node has been manually fenced, allowing Pacemaker to proceed with recovery. It is used when fencing was performed outside the configured STONITH device.

Question 41High Availability Cluster Management

An administrator runs `pcs node standby node1` on a two-node Pacemaker cluster where resources are currently running on node1. What is the immediate effect?

  1. node1 remains active but its resources are stopped and not started elsewhere.
  2. node1 is removed from the Corosync membership permanently.
  3. node1 is fenced and rebooted before resources are moved.
  4. node1 is placed in standby; Pacemaker moves its resources to node2 and will not start new resources on node1. Correct answer.

Explanation

Standby marks a node as ineligible to host resources. Pacemaker evacuates resources from node1 to an eligible peer, node2, and keeps node1 out of placement until it is unstandby. It is not a fencing or permanent removal operation.

Question 42High Availability Cluster Management

Which file contains the primary Corosync cluster configuration, including the nodelist and totem settings?

  1. /etc/pacemaker/cib.xml
  2. /etc/corosync/corosync.conf Correct answer.
  3. /etc/sysconfig/pacemaker
  4. /var/lib/pacemaker/cib/cib.xml

Explanation

Corosync reads /etc/corosync/corosync.conf for membership, nodelist, totem, and quorum settings. Pacemaker's cluster state is stored in the CIB, which is a separate file under Pacemaker's control.

Question 43High Availability Cluster Management

An administrator runs `pcs resource cleanup web_server` after resolving the cause of repeated monitor failures. What does this command do?

  1. It deletes the resource definition and its history.
  2. It disables the resource on all nodes until manually enabled.
  3. It restarts the resource agent process on every node.
  4. It clears the resource's failure count and operation history so the cluster can schedule it afresh. Correct answer.

Explanation

`pcs resource cleanup` resets the resource's fail-count and removes recorded operation history. This allows Pacemaker to attempt the resource again without waiting for failure-timeout or manual clearing of individual failures.

Question 44High Availability Cluster Management

Given the constraint `pcs constraint order promote db-clone then start app`, what behavior does Pacemaker enforce?

  1. app must run on the same node as the promoted db-clone instance.
  2. app and db-clone must be promoted together.
  3. db-clone must be stopped before app can start.
  4. No instance of app may start unless a db-clone instance has been promoted. Correct answer.

Explanation

An order constraint with the `promote` action requires the first resource to reach the promoted role before the second resource may start. It sequences promotion of db-clone ahead of app startup; it does not by itself force colocation.

Question 45High Availability Cluster Management

A resource `vip` has only this location constraint: `pcs constraint location vip prefers node1=100`. Node1 becomes unavailable while node2 is healthy. What does Pacemaker do?

  1. It stops vip and waits for node1 to return.
  2. It fences node2 to prevent vip from running there.
  3. It moves vip to node2 because a finite positive score on node1 does not exclude other nodes. Correct answer.
  4. It moves vip to node2 only after removing the constraint manually.

Explanation

A `prefers` score of 100 is a positive preference for node1, not an infinite constraint. If node1 is unavailable, node2 is still eligible, so the scheduler can place vip there.

Question 46High Availability Cluster Management

An administrator sets `stonith-enabled=false` on a production Pacemaker cluster. Which risk does this introduce?

  1. Corosync will refuse to form a cluster with more than two nodes.
  2. SBD will automatically reboot every node in the cluster.
  3. A failed node may still access shared storage while its resources are started elsewhere, causing data corruption. Correct answer.
  4. Resources cannot be monitored by Pacemaker.

Explanation

Disabling STONITH removes the cluster's ability to confirm that a failed node is powered off or isolated. Without that guarantee, Pacemaker may start resources on another node while the old node can still access the same data, risking corruption.

Question 47High Availability Cluster Management

What is the purpose of a quorum device (qdevice) in a two-node Pacemaker cluster?

  1. It acts as a fencing device to power off an unresponsive node.
  2. It provides an additional vote so the cluster can retain quorum when one node fails. Correct answer.
  3. It replaces Corosync with a Pacemaker-native messaging layer.
  4. It stores the CIB so both nodes always have an identical copy.

Explanation

A quorum device contributes an external vote to the cluster's quorum calculation. In a two-node cluster, this can allow the surviving node to keep quorum when its peer is lost, without requiring a full third cluster node.

Question 48High Availability Cluster Management

In an SBD-based cluster, what is the function of the watchdog timer?

  1. It reboots the node if SBD stops refreshing it, providing self-fencing when the node hangs. Correct answer.
  2. It votes on quorum membership in place of Corosync.
  3. It monitors shared storage latency and moves resources if latency rises.
  4. It restarts the Pacemaker scheduler every 60 seconds.

Explanation

SBD periodically refreshes a hardware or software watchdog while the node is healthy. If the node hangs or SBD cannot run, the watchdog expires and resets the node, which prevents it from continuing to access cluster resources.

Question 49High Availability Cluster Management

A cluster has a fencing level 1 device that fails to fence node1, and a fencing level 2 device is configured for node1. What does Pacemaker do next?

  1. It declares the cluster ineligible for quorum.
  2. It retries the level 1 device indefinitely and never uses level 2.
  3. It abandons fencing and starts node1 resources on node2 immediately.
  4. It attempts the level 2 device; if that succeeds, fencing is considered successful. Correct answer.

Explanation

Pacemaker tries fencing devices by level, moving to the next level only when the current level fails. If the level 2 device succeeds, the node is considered fenced and recovery can proceed.

Question 50High Availability Cluster Management

In `corosync.conf`, which `totem` setting makes Corosync use both redundant rings simultaneously for duplicate traffic rather than keeping one ring as standby?

  1. `rrp_mode: none`
  2. `rrp_mode: active` Correct answer.
  3. `secauth: on`
  4. `rrp_mode: passive`

Explanation

With `rrp_mode: active`, Corosync sends duplicate traffic over all configured rings at the same time. `passive` keeps one ring as a standby, and `none` disables redundant ring operation.