What Does LPIC-3 High Availability Actually Certify?

It certifies that you can keep Linux services running when individual machines do not. The objectives cover load balanced clusters, failover clusters, replicated block storage, clustered and distributed file systems, and the single-node resilience underneath all of it. LPIC-3 is LPI’s enterprise tier, and this is one of four specialties within it.

The distribution-neutral framing matters here more than at the lower tiers. The exam names specific software rather than specific distributions: Pacemaker and corosync for failover, keepalived and HAProxy for load balancing, DRBD for replicated block devices, GlusterFS and Ceph for distributed storage. Those are the tools an enterprise Linux estate actually runs, whichever vendor supplies the operating system.

One of four specialties, not a single ladder rung

LPIC-3 offers four separate specialty certifications, and passing any one of them grants LPIC-3 for that specialty. High Availability and Storage Clusters is the 306 track. There are also mixed environment, security, and virtualization and containerisation tracks, each with its own exam code. You do not work through them in sequence; you choose the one that matches your work.

Why Can You Pass 306-300 and Still Not Hold LPIC-3?

Because LPI requires an active LPIC-2 certification before it awards LPIC-3. Passing 306-300 satisfies the exam requirement, but the certification itself is only granted to a candidate who already holds a current LPIC-2. Without it, the pass sits there unredeemed until the prerequisite is met.

The word doing the work is active. LPI certifications carry a validity period, so an LPIC-2 earned years ago and never renewed does not count. A candidate returning to LPI after a gap needs to check the state of their LPIC-2 before booking anything, not afterwards.

What this means in practice

If you do not hold LPIC-2, the honest sequence is LPIC-1 (exams 101 and 102), then LPIC-2 (exams 201 and 202), then 306-300. That is a substantial commitment, and it is the reason this credential sits where it does: the LPIC-2 requirement is what makes LPIC-3 an enterprise-level statement rather than a topic badge.

It is worth being blunt that several published summaries of 306-300 list no prerequisite at all. Checking the specialty against the 306-300 exam page alongside LPI’s own certification page is the way to avoid an expensive misunderstanding.

Where Do the Sixty Marks Sit Across Twelve Objectives?

LPI publishes a weight for every objective, and the twelve weights sum to 60, matching the question count. That makes the blueprint unusually transparent: a weight of 8 means roughly eight questions. Cluster Management takes 22, Cluster Storage 13, Distributed Storage 13, and Single Node High Availability 12.

Three objectives carry a weight of 8 and therefore matter disproportionately: Load Balanced Clusters, Failover Clusters and Ceph Storage Clusters. Together they are 24 of 60, roughly 40 percent of the paper across three topics. At the other end, Hardware and Resource High Availability and Advanced RAID are worth 2 each, so four questions between them.

What Does the Clustering Half of the Exam Cover?

Cluster Management is the heaviest topic area at 22 of 60. It splits into concepts and theory at weight 6, load balanced clusters at 8, and failover clusters at 8. The theory objective sets up the vocabulary the other two assume: active/passive against active/active, failover against load balanced, shared-nothing against shared-disk.

Load balanced clusters, weight 8

This objective names LVS and IPVS, VRRP, keepalived, ldirectord, backend server networking, and HAProxy configuration. It is a mix of kernel-level load balancing and userspace proxying, and the exam expects configuration familiarity rather than architectural description alone. HAProxy in particular appears both as understanding and as configuration.

Failover clusters, weight 8

Pacemaker dominates: managing cluster configurations and resources, obtaining node information and managing node health, quorum and fencing, the split brain detector on shared storage, and managing Pacemaker through both pcs and crmsh. Pacemaker ACL awareness is included too.

Fencing and quorum are the concepts that separate candidates who have run a cluster from those who have read about one. A cluster that cannot fence a misbehaving node cannot safely recover from a partition, and the exam returns to that idea repeatedly. Knowing both pcs and crmsh matters because different distributions ship different tooling around the same Pacemaker core.

Why Is Distributed Storage Worth as Much as Failover?

High Availability Distributed Storage carries 13 of 60, the same as Cluster Storage, and Ceph alone is weighted 8, equal to the two heaviest clustering objectives. GlusterFS takes the remaining 5. LPI weights it that way because distributed storage is where modern Linux high availability has actually moved.

The reasoning is architectural. A traditional failover cluster keeps a service alive by moving it between nodes attached to shared storage, which makes the storage itself the thing that must not fail. A distributed store spreads data across many nodes so that losing one is routine rather than exceptional. Understanding how Ceph is structured is the single highest-value piece of study on this exam by weight.

DRBD sits between the two models

DRBD is its own objective at weight 6, covering architecture, resources, states and replication modes, disk and device configuration, networking connections and meshes, automatic recovery and error handling, drbdadm, the principles of drbdsetup and drbdmeta, integration with Pacemaker, and the architecture of LINSTOR. Protocols A, B and C and the primary and secondary roles are named explicitly.

DRBD is replicated block storage rather than a distributed file system, which makes it the bridge between the failover model and the distributed one. Its integration with Pacemaker is why it appears in an exam that also covers Ceph: in practice they solve overlapping problems at different layers.

What Does Single Node High Availability Add?

Single Node High Availability is the lightest topic area at 12 of 60, covering hardware and resource availability at weight 2, advanced RAID at 2, advanced LVM at 3, and network high availability at 5. It is the resilience you build into one machine before any clustering happens.

The low weights are easy to misread as unimportance. They are better read as a signal about how few questions each is worth: two questions on RAID, two on hardware and resource monitoring. That is a small return on deep study, which makes this the area to cover competently and quickly rather than exhaustively.

Network high availability is the exception

At weight 5, network high availability is worth more than the other three single-node objectives combined would suggest, and it connects directly upward to the load balancing material. Bonding, redundant paths and interface-level failover are the foundation the cluster layer assumes is already in place.

Why the Linux Foundation context helps

Almost every technology named across these objectives is an open-source project rather than a vendor product, and several sit under shared governance. Browsing the Linux Foundation project landscape is a quick way to see how these components relate to one another, which helps when the exam asks about integration rather than about a single tool.

What Is the Exam Format, and How Long Does It Last?

306-300 is a 90-minute exam of 60 multiple-choice and fill-in-the-blank questions, with a passing score of 500 out of 800. It costs around $200 USD depending on region, and it can be taken either at a Pearson VUE test centre or online through OnVUE. The certification is valid for five years.

Fill in the blank changes how you study

A multiple-choice question lets you recognise the right command. A fill-in-the-blank question makes you produce it. Given that the objectives name drbdadm, pcs, crmsh, lvresize, lvchange and the gluster subcommands explicitly, the difference is not academic. Recognition-level familiarity with the tooling is not enough for this format.

The current version and what it replaced

The objectives are version 3.0, which came out of a split of version 2.0 of the older 304 exam. The previous version, 304-200, was available only until June 2022, so any study material written against 304 is describing a different, narrower syllabus. Checking the version on any resource before using it is worth the thirty seconds.

How Should You Work Through the Four Topic Areas?

Follow the weights. Cluster Management is 22 of 60 and Ceph alone is 8, so those two claims on your time come first. Single Node High Availability is 12 of 60 across four objectives, so cover it competently and move on. Confirm your LPIC-2 is active before you plan anything at all.

1. Verify that your LPIC-2 certification is current, because without it a pass on 306-300 will not result in an LPIC-3 certification.
2. Start with high availability concepts and theory, since the vocabulary of active/passive, shared-nothing and quorum underpins every heavier objective that follows.
3. Work failover clusters and Pacemaker next, practising with both pcs and crmsh, and treating fencing and quorum as the ideas the exam keeps returning to.
4. Take Ceph after that, because at weight 8 it is worth as much as either clustering objective and is usually the least familiar to candidates from a traditional cluster background.
5. Finish with load balancing, DRBD and the single node objectives, then drill the named commands until you can produce them from memory rather than recognise them.

That last step is the one specific to this exam’s format. Fill-in-the-blank questions on drbdadm, pcs, crmsh, lvresize and the gluster subcommands reward hands-on repetition and punish reading. Building a small cluster in virtual machines is worth more than any amount of note-taking, and the practical emphasis is a recurring theme in these 306-300 preparation notes.

If you are choosing between LPIC-3 specialties

The four tracks are genuinely different jobs, not difficulty tiers, and picking on interest rather than on what your estate runs is the common mistake. The mixed environment track suits an administrator whose reality is Linux alongside Windows, and the 300-300 route covers that side of the choice.

Frequently Asked Questions

Do you need LPIC-2 before taking 306-300?

You need an active LPIC-2 certification to be awarded LPIC-3. You can sit and pass 306-300 without it, but the certification will not be granted until the prerequisite is met.

How many questions are on the LPIC-3 306 exam?

60 questions in 90 minutes, in a mix of multiple-choice and fill-in-the-blank formats.

What is the passing score for 306-300?

500 out of 800.

How much does LPIC-3 High Availability cost?

Around $200 USD. LPI prices the exam by country, so the exact figure depends on where you sit it.

How long is LPIC-3 valid?

Five years, with options available for extending beyond that period.

Can you take 306-300 online?

Yes. It is available at Pearson VUE test centres and online through OnVUE, in English and Japanese.

Which objective is worth the most marks?

Three are tied at weight 8: Load Balanced Clusters, Failover Clusters and Ceph Storage Clusters. Together they account for 24 of the 60 marks.

What software does the 306 exam actually name?

Pacemaker with pcs and crmsh, keepalived, ldirectord, HAProxy, LVS and IPVS, DRBD with drbdadm and LINSTOR, GlusterFS, Ceph, and LVM and RAID tooling.

How many LPIC-3 specialties are there?

Four. Passing any one of the four exams grants LPIC-3 for that specialty, so they are parallel tracks rather than sequential levels.

Is study material written for exam 304 still usable?

Not reliably. The current objectives are version 3.0 and came from a split of version 2.0 of exam 304, whose last availability was June 2022. Material written against 304 describes a different syllabus.

Conclusion

LPIC-3 High Availability is a genuine enterprise Linux credential, and its blueprint is more honest than most: twelve objectives with published weights that sum exactly to the sixty questions, so you can see precisely where the marks are. Cluster Management takes 22, Ceph alone takes 8, and the whole Single Node area is worth 12.

The detail to act on first is not in the objectives at all. LPI awards LPIC-3 only to candidates holding an active LPIC-2, and that condition is missing from a number of published summaries of this exam. Confirm it before booking, then follow the weights: theory, failover clustering, Ceph, and finally the commands you will have to type from memory rather than recognise.