Almost KIA'd everything...

[danielcoatesdanielcoates is verified member.]

Ok, so I decided to change the scheama that runs the site, this resuled in a mad scramble to fix it after it just killed everything, how often have any of you done anything like that, thankfully we recovered in about 8 hours, and there's been minimal data loss on active dropping domains, just some history, and it led to some major discoveries and breakthroughs.

 

[Invest93]

Why did you want to change the schema?

 

[danielcoatesdanielcoates is verified member.]

Why did you want to change the schema?

It was using a large number of joins, resulting in slow loading of the site >1,200ms, the changes made the site load in <400ms when it’s warm

 

[Invest93]

What schema was you using?

And how did it get to that much?

 

[danielcoatesdanielcoates is verified member.]

We are using Postgres, it wasn't so much the volume of data, but that it was normalised, and running through 10M+ domains, with 4-5 joins, to solve the issue we essentially join the data for the main droplist table into a de-normalised single table, with only 2 joins if a user looks at the information panel, this is lazy loaded, so only loads when it needs to, rather than on every page load. - There are still slow-down issues due to the way the database works, if nobody is using the site, it goes to sleep, and takes a second or two to wake back up, but once that initial load is done, speeds are back down to under 300ms.