Nomad Associate - Exam Strategy¶
60 minutes, 57 questions, multiple-choice and multi-select. About 63 seconds per question on average. Pace is important; lingering on one hard question costs you two easy ones.
The Week Before¶
- Run PSI system check 48+ hours out.
- Confirm testing space is quiet and clutter-free.
- Normal sleep; don't cram.
- One full timed mock, review gaps.
The Day Of¶
- Real meal 60-90 min out.
- Close all apps and notifications.
- Bathroom break right before check-in.
- Water nearby.
Time Budget¶
- 57 questions / 60 minutes = 63 sec/q average
- First pass: ~45 minutes at 47 sec/q; flag and skip hard ones
- Second pass: 12 minutes on flagged
- Final 3 minutes: verify all answered
Question Approach¶
- Read carefully. Identify the ask.
- Identify the domain (arch, jobspec, scheduling, networking, storage, ACL).
- Eliminate clearly wrong answers.
- Pick the most specific of the remaining.
- Flag if unsure, move on.
Common Trap Patterns¶
Job Type Confusion¶
service: long-running, auto-restart, scalablebatch: run to completion, not restartedsystem: one allocation per node in scope (like K8s DaemonSet)sysbatch: run to completion on every node (one-off cluster ops)
Questions describe behavior, ask for the type. Know all four.
Driver Selection¶
dockerfor container imagesexecfor chrooted/cgrouped native binariesraw_execfor untrusted/unsandboxed (disabled by default)javafor JAR filesqemufor KVM VMspodman,containerdas Docker alternatives
Watch for wrong driver assignments (e.g., "java" for a JAR is correct; "exec" is wrong though technically possible).
Network Modes¶
host: task shares host network; ports are host portsbridge: Nomad creates a network namespace with NATnone: no network- CNI plugins: custom (calico, etc.)
bridge is common and supports port mapping. host is high-performance but no isolation.
Service Provider¶
provider = "consul"(default): registers with Consulprovider = "nomad"(1.3+): Nomad native catalog
Some questions ask which requires Consul installed. Only the consul provider does.
Constraints vs Affinities¶
- Constraints: hard requirements; fail to place if unmet
- Affinities: soft preferences with weights; scheduler prefers but doesn't require
Questions describing "must" or "required" want constraints. "Prefer" or "should" wants affinities.
Canary vs Rolling Update¶
Canary: N new-version instances run alongside old. Promote manually (or via auto_promote) to replace all. Rolling: no overlap; each old instance replaced one at a time.
Restart vs Reschedule¶
- restart: task stays on same alloc, Nomad restarts the task
- reschedule: alloc itself moves (to a new node) after restart attempts exhausted
Restart first, then reschedule.
Raft Quorum Math¶
- 3 servers: quorum 2, tolerates 1 failure
- 5 servers: quorum 3, tolerates 2 failures
- 7 servers: quorum 4, tolerates 3 failures
- Recommended: 3 for small, 5 for large. Never use even numbers.
System Jobs Don't Honor count¶
system and sysbatch jobs run one allocation per node matching constraints. count is ignored or implied.
Default Ports¶
- 4646: HTTP API and UI
- 4647: RPC
- 4648: Serf gossip
Memorize these.
Namespace vs Region¶
- Namespace: logical partition for multi-tenancy
- Region: physical grouping (independent servers)
A job belongs to one region and one namespace.
Frequently Tested Topics¶
- Which files are auto-loaded? (Nomad reads
-configdirectory recursively) - What does
nomad job plando? (Dry run, shows diff and scheduling result) - What's the difference between
-purgeand regular stop? (Purge removes job spec from state entirely) - How do periodic jobs work? (Cron-like
periodic { cron = "..." }) - How does service mesh work with Consul Connect? (Sidecar proxy injected per task)
- When is preemption used? (Higher-priority jobs can evict lower-priority)
Answer Patterns That Are Often Wrong¶
- Options mentioning Kubernetes concepts (kubelet, pod, ingress controller directly) in Nomad context
- Options suggesting YAML as Nomad's config (HCL and JSON only)
- Options claiming
raw_execis safe default (it's disabled by default) - Options confusing system with service jobs
Answer Patterns That Are Often Right¶
- Options specifying "HCL" or "HCL2" for config format
- Options mentioning Consul for service discovery by default
- Options highlighting the scheduler's bin-packing behavior
- Options referencing Raft for server consensus
Multi-Select Considerations¶
Some questions ask for all correct answers. Evaluate each option independently. Typically 2-3 correct out of 4-5.
Reading Difficult Questions¶
Focus on:
- Goal: what outcome is desired
- Constraint: what must/must not happen
- Environment: OSS vs Enterprise, with/without Consul, with/without Vault
Mental State¶
- This is Associate-level. You studied; trust it.
- Flag and move on when stuck; fresh eyes help on pass 2.
- Don't change confident answers unless you see a specific error.
After Submitting¶
- Results often within minutes.
- Credly badge email for pass.
- Fail? Review domain breakdown for retake focus.
Post-Certification¶
- Update LinkedIn and resume.
- Deploy a real Nomad cluster at work or home to cement skills.
- Consider Consul Associate next.