I've been a SQL Server DBA for the better part of a decade, mostly on-prem, and the last two years I've been dragged kicking and screaming into Azure. DP-300 (Administering Microsoft Azure SQL Solutions) had been sitting on my "should probably do that" list forever. In June my manager finally tied a small bonus to it, so I gave myself three weeks and kept a rough log of what I did each block. This is that log, cleaned up a bit. Maybe it saves you some flailing.
Quick note before the day-by-day: the thing that made the biggest difference wasn't a course. It was hammering a free DP-300 practice test early and often so I could see where my on-prem instincts were wrong. More on that below.
Week 1 — Find out how wrong you are
Block 1 (Sat morning). I did not read anything. I went straight to a practice set cold, no studying, just to get a baseline. Scored badly, obviously, but that was the point. The wrong answers told me exactly which topics my SQL Server background did not cover: purchasing models (DTU vs vCore), the difference between Azure SQL Database, Managed Instance, and SQL on a VM, and the whole elastic pool concept. If you're coming from on-prem like me, that PaaS-vs-IaaS decision tree is where you'll bleed points.
Block 2 (Sat afternoon). Read the official Microsoft Learn "Administer a SQL Server database infrastructure" and "Deploy and configure Azure SQL" paths. I skimmed the stuff I already knew (indexes, statistics, query plans — that transfers straight over) and slowed way down on deployment options. I made a one-page cheat sheet: when do you pick Managed Instance vs single database vs VM. Cross-database queries and SQL Agent? That pushes you toward MI. Full OS control or unsupported features? VM. Wrote it in my own words, which forced me to actually understand it.
Block 3 (Sun). Spun up an actual Azure SQL Database and a Managed Instance in a throwaway subscription. Do this. Reading about "the Managed Instance takes ~4 hours to deploy" is nothing like watching your portal blade spin for the whole afternoon and understanding why that matters for a migration plan. I configured a firewall rule, connected from SSMS, locked myself out, fixed it. That mistake stuck harder than any flashcard.
Block 4 (Sun evening). Security. Azure AD (Entra) authentication, contained users, Transparent Data Encryption with customer-managed keys, Always Encrypted, dynamic data masking, row-level security. This is heavily weighted on the exam. I set up an Entra admin on my test DB and created a contained database user, because "CREATE USER ... FROM EXTERNAL PROVIDER" is the kind of syntax detail they love to test.
By the end of week 1 I retook a practice set. Went from embarrassing to merely bad. Progress.
Week 2 — The boring, heavily-tested middle
This week was HA/DR and automation, which is where most of the exam weight actually lives.
Block 5. High availability and disaster recovery. Active geo-replication vs auto-failover groups vs the built-in zone-redundant config. I drew the topology for each on paper. The exam wants you to know which one gives you a read-write listener endpoint that survives failover (auto-failover groups) versus which one you have to manually repoint (geo-replication). Also backup retention: the difference between the automatic 7–35 day PITR window and long-term retention policies. Know the defaults.
Block 6. Monitoring. Query Store, Extended Events, Azure Monitor, sys.dm_db_* DMVs, Intelligent Insights, automatic tuning. I turned Query Store on and deliberately wrote a bad query to watch a plan regression show up. The exam loves scenario questions like "queries suddenly got slow after a deployment, what do you check first" — and the answer is almost always Query Store forced plans.
Block 7. Automation. Elastic jobs (the Azure SQL Database replacement for SQL Agent, since single databases don't have Agent), Azure Automation runbooks, and PowerShell/az cli for scaling. I wrote a tiny runbook to scale a database up on a schedule. Even if you never do this at work, seeing the shape of it makes the multiple-choice options obvious.
Block 8. This is where I leaned on the DP-300 exam page to make sure I'd actually covered the whole objective domain and wasn't just studying my favorite topics. I checked my cheat sheet against the skills-measured list and found two gaps I'd skipped entirely: sp_configure-style server config on Managed Instance, and resource governance. Fixed both.
Retook a full-length practice set at the end of the week. Passing range now, but shaky on HA/DR wording.
Week 3 — Drill the wording, not the concepts
At this point I knew the material. The remaining problem was Microsoft's question style: long scenarios, "you need to minimize cost" or "minimize downtime" or "minimize administrative effort" — same tech, different constraint, different correct answer.
Block 9–10. Pure question drilling. I did sets of 20–30, then — and this is the part that mattered — I read the explanation for every single one, including the ones I got right, to check I got them right for the reason and not by luck. When I couldn't articulate why the other three options were wrong, I went back to Learn for that topic. For the trickier scenario questions I ran them through the AI simulator at ai.examcert.app because it gives instant explanations and I could ask follow-up "but what if it were a Managed Instance instead" questions, which is exactly how the exam mutates a scenario.
Block 11. Weak-area cleanup. Mine were auto-failover group endpoints and TDE with customer-managed keys in Key Vault. I re-did those specific labs.
Block 12 (day before). Light. One timed practice set to confirm pacing — you get roughly a couple minutes per question and the long scenarios eat that fast, so practice skipping and flagging. Then I stopped. No cramming the morning of.
What I'd tell past-me
- Don't trust your on-prem instincts. The PaaS deployment/HA/DR decisions are the whole game and they're genuinely different.
- Lab the security and failover stuff with your own hands. Reading isn't enough for those.
- Practice questions aren't a final check — they're your map. Start them day one to find the holes, not week three to confirm you're done.
- Read wrong-answer explanations religiously. On DP-300 the constraint word ("cheapest", "least downtime") is the whole question.
Passed it. The bonus cleared. Now I actually reach for auto-failover groups at work without second-guessing, which is worth more than the cert honestly.
Top comments (0)