Ask HN: What are must know DBA concepts for developers? | Hacker News Reader

Ask HN: What are must know DBA concepts for developers?

Could you tag your responses as beginner/intermediate/advanced concepts?

Understand ACID and the isolation levels available in your database server.

Understand transactions.

Read an explain plan and understand basic indexing.

Test recovery of your server and implement a standby/ failover.

Understand the pros and cons of where to put the logic of the application (db objects, stored proc, backend services, ui etc).

Understand the pros and cons of using an orm and how you might tune or scale an app using one.

Do you have any pointers as to where one might learn more about the topics you mentioned? I'd be interested in reading more about #5 in particular

One of the problems is that people will give different answers depending on their expertise and there are very few people who have deep knowledge of database development and backend development and supporting large production systems because in the larger enterprise organizations people tend to specialise. I have around 20 years experience using Oracle database and can kind of translate that to other database servers e.g. postgres, sql server etc. I have a bit of knowledge of writing services in c#, ruby or java. Previously I would have advocated putting all the logic into the database but I'm starting to realise some of the compromises involved.

Reading the Oracle or postgres documentation gives a lot of information about the internals, tuning, acid etc although it can be a bit dry...

That's an interesting perspective. Coming from a background of primarily developing backend services and using databases as a necessary tool, I usually approach it the other way and steer clear of stored procedures, triggers and the like. I guess it's also a matter of using the tools that you're most comfortable with.

Of course. At least in the UK and probably elsewhere large businesses, banks, government etc often have a large centralized systems based on a single or set of relational databases. A lot of the logic was controlled on the database either by constraints/checks etc or stored procedure access so it didn't have to to be replicated across the many disparate systems which connect to it. This is slowly changing, very slowly in some cases!

I can understand why that might be a surprise if you've been working for <10 years and mainly on new SAAS style products.

On the deployment side of things...

Many developers don't get the concept of deploying a patch. They envision a script for each DB object stored in source control. Mimicking the the Java class files for all their objects. That's fine for code which can be re-built from scratch. But you can't re-build your DB, it must be patched to preserve the precious data.

It works fine for DB release #1. But when DB release #2 comes around there's no way to deploy the DB changes. Special diff tools are brought in to generate the patch. The auto generated patch drops a column.. ooops.

intermediate/advanced

Know that most databases can spit out a "worst performing queries" and "overall most expensive" reports. The method varies per engine. Always illuminating, and usually not the queries you were expecting.

Know how to turn logging on/off and where the logs go.

Know there are a bunch of system tables that keep data about all your tables in it in most engines, information_schema tables. Rows, size, columns, etc. These are actually defined in the SQL spec, but various SQL vendors have even more available, e.g. there's a load of extra stored procedures in SQL Server.

Be able to generate and read a query plan report for your engine, which will tell you what's causing a slow query.

Reply on news.ycombinator.com