Episodes / Episode 3 of 9
DBA Pushback & Project Kickoffs: Say What It Costs
DBAs get invited to projects at the end, after every decision that matters is made. How to get invited early, what to ask at the kickoff, and how to push back by saying what it costs.
Watch on YouTubeFull course playlistSubscribe
In this episode
- Get invited to the kickoff, not the go-live
- Six kickoff questions: RTO and RPO, growth and retention, OLTP vs OLAP, licensing, CIS, CNAMEs
- Pushback: your job isn’t to say no, it’s to say what it costs
- The priority list, and the three ways every pushback ends
- Advising: give one recommendation, not five options
Chapters
Transcript
Read the full transcript
Get invited early
DBAs get invited to projects at the end. “We go live Friday. Can you build us a server?” By then, every decision that matters is already made, and half of them are wrong.
Ask to be at the kickoff. Not the status meetings. Not the weekly sync. The kickoff. Thirty minutes of you at the start saves thirty weekends of you at the end.
I have been called into an incident where a sensitive OLTP application kept going down. Instead of building dedicated SQL Servers, or using an existing OLTP SQL Server, they put the databases on an OLAP SQL Server that ran heavy reporting workloads. Whenever an OLAP workload ran, it brought down the OLTP application. The customer-facing outages and the migration work effort could have been easily avoided by a DBA being involved early.
The kickoff questions
Bring the same list to every project kickoff. Every one of these questions is a whole episode of this course, which is not a coincidence.
How long can this be down, and how much data can we lose? That’s RTO and RPO: the recovery time objective and the recovery point objective. Episode 6. If they say “zero and zero,” you’ll get to tell them what that costs, and they can’t afford it.
How big is it, and how fast will it grow? And what’s the retention plan? An app that goes live with no retention is how a small, fast database turns into a huge, slow one. That’s episode 8. Seconds or minutes? If people get angry when it doesn’t answer in seconds, it’s OLTP, online transaction processing. If they get angry when it doesn’t answer in minutes, it’s OLAP, online analytical processing. They get built differently. Episode 8 again.
Who is paying for the SQL Server licenses, and which edition? Episode 4. Ask this one early, because the answer changes the hardware. What in the CIS Benchmark will this app break? Episode 7. Find out now, not during the security audit. Does the connection string use a CNAME? If it doesn’t, it will, before go-live. Episode 8.
If nobody in the room can answer these, the project isn’t ready to go live. It’s ready to go wrong.
Pushback
Your job isn’t to say no. Your job is to say what it costs. When you say the sun will rise, you’re not commanding it to rise. You’re informing everyone of what will happen. If you’re not there, the sun will still rise. It just might be a surprise.
Keep a running list of what you’re working on. When someone hands you something new, say “I can do it. Which one of these doesn’t happen?” Now it’s their decision, and it should be.
Every pushback conversation ends in one of three places.
The boss accepts the risk. The developers change the app. You add hardware.
If it ends with the boss accepting the risk, get it in an email. Not because you don’t trust your boss. Because in a year, when the risk shows up, nobody will remember the meeting, and you’ll be the one explaining it.
You can always throw hardware at bad code. That’s a real option, and sometimes it’s the right one. More RAM is usually cheaper than a month of developer time, but not always. Price it out. More CPU cores are a different story, because every core needs a SQL Server license. Put a number on it, and let the people with the budget pick.
Advising
You’re the expert in the room. Act like it.
Give one recommendation, not five options. Five options sounds like you don’t know. One recommendation sounds like you do. Say something smooth like, “I recommend X. If we don’t do X, here’s what happens, and here’s about when.”
If they don’t take your advice, that’s OK. It’s their app and their budget. Write it down, and move on.
And make friends with the developers. If the developers don’t like you, they will fight you on every one of these. If they do like you, they’ll come ask the kickoff questions before you do.
