10 exam-style questions with answers and explanations, straight from our 1,030-question bank. Tap an answer to check yourself. When you're ready, take the scored version in the free practice test.
These 10 free High Availability Clustering questions are organized by exam domain, so you can see how each part of the Red Hat Certified Specialist in High Availability Clustering blueprint is tested. Reveal the answer and explanation under each question.
A newly configured RHEL 8 cluster works after pcs cluster start --all. Before a reboot test, its daemon status reads: corosync: active/disabled pacemaker: active/disabled pcsd: active/enabled After reboot, node authentication still works, but Corosync and Pacemaker are not running. The configuration files are intact. Which omitted command addresses this reboot-only failure?
Correct answer: A - pcs cluster enable --all
Seven cluster nodes each have one vote, and expected_votes remains fixed at 7. After all seven have joined, a communication failure splits them into groups of five and two. There is no quorum device, tie-breaker, two-node mode, or dynamic quorum adjustment. Pacemaker has no-quorum-policy=stop. An ordinary failover service is running on the two-node side and can stop cleanly. What should occur on that side of the partition?
Correct answer: C - The service stops there; only the five-node partition retains quorum.
A shared-storage database was running on node-c when cluster communication with that node was lost. The two surviving nodes retain quorum, but fencing through fence_ipmilan has timed out. No one knows whether node-c is still writing to the database volume. An operator proposes running pcs stonith confirm node-c to release the blocked recovery. What evidence must exist before that confirmation can truthfully be issued?
Correct answer: C - Independent confirmation that node-c has actually been powered off.
An email-notification script is moved from a ClusterMon resource configured to run as root to a native Pacemaker alert. The script now reads the CRM_alert_* environment variables correctly. Its recipient, executable permissions, dependencies, and SMTP connectivity have been verified. It still sends mail when tested as root, but alert invocations fail while opening /etc/cluster/mail.secret. That file is owned by root:root with mode 0600; its parent directories are searchable and SELinux permits access. Which correction addresses the execution-context change without making the credential public?
Correct answer: D - Give the hacluster account read access to the protected credential file.
An active/passive web service uses a Pacemaker group ordered as follows: volume-group activation, filesystem mount, floating IP, Apache. The filesystem cannot start on either eligible node. The volume-group resource remains Started, while the IP and Apache remain Stopped. An administrator suspects that the group has failed to roll back correctly. How should this state be interpreted?
Correct answer: B - Earlier members may remain started; later members are blocked by the failed dependency.
After a controlled web-service move, a request sent directly to the floating IP returns an old maintenance page. The former owner has stopped Apache and released the IP; web-b now owns the IP and answers the request. The effective DocumentRoot on both nodes is /srv/portal. Before the move, that directory was a mount of /dev/vg_portal/lv_content. On web-b, findmnt -T /srv/portal instead identifies /dev/mapper/rhel-root mounted at /. Which diagnosis best fits all these findings?
Correct answer: B - Apache is serving the underlying local directory because the shared filesystem is not mounted.
The billing resource is a single primitive running on node-b. Its effective resource-stickiness is 300. Location constraints give it +500 on node-a and +100 on node-b. There are no other score contributions, failures, capacity limits, or placement restrictions. Both nodes are healthy, and node-a has just been taken out of standby. At the next scheduling decision, Pacemaker should:
Correct answer: A - Move billing to node-a: its score is 500, compared with 400 on node-b.
A resource has migration-threshold=4 and failure-timeout=0. The cluster explicitly sets start-failure-is-fatal=true. Its first start attempt on node-b returns OCF_ERR_GENERIC (1), with on-fail=restart, and Pacemaker makes node-b ineligible immediately. No monitor failures preceded the start attempt. Why did the resource not receive three more start attempts on that node?
Correct answer: A - With start-failure-is-fatal=true, the start failure sets node-b's failcount to INFINITY, bypassing its threshold of four.
During a storage-fabric incident, an application continues successful reads and writes through /dev/mapper/orders_data. The map has two paths to the same LUN: sdb: failed faulty running sdc: active ready running The paths use independent fabrics, multipathd is running, and the peer cluster node has lost the same fabric. What response is warranted by this evidence?
Correct answer: D - Keep the application online through the usable path while restoring the failed fabric.
A GFS2 filesystem mounted at /srv/projects contains journal0, journal1, and journal2. Three nodes currently mount it read/write; two additional nodes will also mount it read/write, without the spectator option. Fencing and locking are healthy, and the filesystem has ample free space. The administrator wants exactly five journals and no service outage. Which operation meets that requirement?
Correct answer: B - Run gfs2_jadd -j 2 /srv/projects once on a node that already mounts the filesystem.
The High Availability Clustering exam also covers these domains. Drill them in the full free practice test:
The full bank has 1,020 more High Availability Clustering questions with explanations.
Continue in the free practice test →
View plans