LPIC-3 High Availability and Storage Clusters study plan

Install Corosync and Pacemaker on three nodes, authenticate with pcs, and bring the cluster online with a unique cluster name.

Trace the Pacemaker architecture (CIB, controller, LRM, stonithd, pengine) and explain which daemon owns each piece of state.

Configure votequorum for a two-node cluster and add a quorum device with pcs quorum device add, then verify with corosync-quorumtool.

Use pcs status, crm_mon, crm_verify and journalctl to classify at least five common cluster start-up failures.

Configure a fence device (IPMI, SBD or hypervisor agent) and prove that a fenced node is rebooted and its resources are recovered.

Create resource groups and primitives for a virtual IP and a service, then test start, stop and migrate operations.

Write ordering, colocation and anti-colocation constraints and verify the resulting placement with crm_simulate.

Place a node in standby and maintenance mode and record the difference in resource behavior.

Build clone and promotable (master/slave) resources and demonstrate controlled promotion and demotion.

Configure booth tickets or a similar site-failover mechanism and simulate a two-site takeover.

Add a pacemaker-remote node and confirm it can host a resource.

Diagnose a split-brain and a resource-stickiness problem using crm_report and CIB diffs.

Export an iSCSI LUN from a target and connect two initiators, confirming consistent device names with udev rules.

Configure device-mapper multipath with two paths and test path failover by disabling one link.

Verify that the cluster fences the storage path owner before a shared resource is started.

Document the interaction between fencing, multipath and shared LUN reservation.

Configure clustered locking with lvmlockd and DLM, create a shared volume group and activate it on two nodes.

Format a GFS2 filesystem, mount it concurrently on two nodes and run a concurrent write test.

Add the filesystem as a cluster resource and verify recovery after a node is fenced.

Configure CTDB-backed clustered Samba or an NFS export served from the cluster and test client access during failover.

Bootstrap a Ceph cluster with cephadm, deploy at least three monitors and six OSDs, and reach HEALTH_OK.

Create pools, set PG counts and explain CRUSH placement rules for the failure domain.

Create and map an RBD image, then mount CephFS on a client.

Simulate an OSD failure and confirm recovery with ceph -s and ceph osd tree.

Build a replicated GlusterFS volume and a DRBD resource, then demonstrate promotion and resynchronization after a node restart.

Configure network bonding or teaming with two slaves and test link failover without losing the management connection.

Assemble a software RAID set, replace a failed member and verify rebuild completion.

Enable a watchdog and kdump, trigger a crash and confirm the vmcore is written.

Configure keepalived VRRP for a floating IP and test takeover by stopping the master.

Review kpatch or live-patching workflow, boot-time service ordering and systemd failure handling.

Re-run the two weakest lab scenarios from weeks 1-7 without notes.

Take a timed 90-question mock exam and review every wrong answer against the relevant domain.