oops this should have been a thread all along
Notice I said in production. I have cause issues on dev before. I have a client that has 700+ sql server databases for a single product. Used to be on only one cluster. To fix a problem across databases manually to each one would take forever. Query took 10 minutes. This was at the end of an outage that was hours and customers were on the war path. I think the non-controversial usage is dev. If it is really bad to do unscripted updates - there would not be a grid editor
2 Replies
and by Query you mean a SQL UPDATE statement? I'm still a little fuzzy on what tool was doing a good job
My workflow would be:
1. Check out the version of code most recently pushed (<1 min)
2. Write a migration (< 5 min)
3. Deploy the backend code, not a full frontend build (<1 min)
4. Run it once in dry run mode to sanity check what it's doing (< 1 min)
5. Run it over all the data (? min)
6. Commit the change to the repo as a reference for future outages + record keeping for postmortem
but obv you're free to do whatever you want. Just sharing my perspective from the experiences I've had & cultural norms, mostly at Dropbox where things were especially oriented around correctness & stability, vs. somewhere like Facebook where I know folks have a lot more leeway
i'm off, hope this was helpful!
Much of this depends on the size of the company. I still believe updates are commonly done during development and are useful. Not critical, just reduces friction
I mean a sql update, sql delete, sql insert.