Take the 2-minute tour ×
Database Administrators Stack Exchange is a question and answer site for database professionals who wish to improve their database skills and learn from others in the community. It's 100% free, no registration required.

I "inherited" a web application which is designed and implemented horribly (both the application and the database). For example, the main data is stored using a sort of emulated key-value storage in a Postgres 8.2 database, making it virtually impossible to extract useful data from it in a reasonable amount of time.

Currently I'm working hard on replacing the entire application + database, however it will take a few months before the new application is finished. This is a problem since the website is really slow due to the extremely bad database design and even worse queries. Therefore I'm planning to fix the database design and the queries on the live site until the website has an acceptable load time as a temporary solution.

I do however have a few limitations to work around, the most problematic ones are:

  • I'll have to do everything from inside a proprietary build CMS.
  • The queries are all distributed over many many files, and there's no way to search for them. Therefore it's next to impossible to guarantee that I've updated all queries to a particular table.
  • I don't have direct access to the database, but I can execute queries using a query editor in the CMS, but only to tables owned by the application (so no pg_* tables for me).
  • I don't have access to a developer environment, nor can I create one. Everything has to be done live.

So basically I'll have to do a migrate a database live over the course of a few days, while simultaneously updating the application live such that it can handle the new application, without being able to search for the usage of each table.

With all these handicaps taken into consideration I came up with the following plan:

  1. Create initially empty tables to store the data in a sane way.
  2. Create triggers on the old tables which can sync the data with the new ones.
  3. Export the data from the old tables to the new tables & enable the triggers.
  4. Replace the queries one by one.

Using this strategy I will have 2 synced databases after step 3. Initially all queries will go to the old database, but while I'm updating the queries slowly the old database will be used less, and the new one more.

Taking this "sub-optimal" situation in mind, is this a good strategy to fix some of the problems? What things should I take into consideration while doing this?

Note that I fully understand that this is a very risky suicide mission. However, I'll have to do something in a short amount of time, otherwise the website becomes entirely unusable.

share|improve this question
    
I am curious how can one limit the access to the system catalog. On the other hand, your plan sounds OK for me after a fast glance, but I would seriously consider leaving that CMS alone on the long run and use/build something friendlier. –  dezso Sep 23 '13 at 14:20
    
The query editor simply analyses the query before executing. Whenever I access a table without a certain prefix, it complains. Also, the entire stack is going to be replaced, so in a few months time I won't have to use the CMS anymore:) –  Tiddo Sep 23 '13 at 14:24

1 Answer 1

up vote 1 down vote accepted

If you can, it would be better to do views and triggers rather than bilateral replication of data between both sides.

So I would modify your plan slightly:

  1. Create a sane schema

  2. Create an emulation of the old schema using views.

  3. Create triggers on the views to write to the new schema instead of the old.

The question is what comes after this. Since you are building a new tool, it is good to be in maintenance mode. This means, I think, only fixing the slowest of the queries that cause you problems.

share|improve this answer
    
I do already use views a lot, but views don't improve performance since they're merely stored queries. Therefore I need to do something else to solve these performance issues. Also, replication will only be one way (old->new), so I guess that makes the operation a little bit easier. I won't change the inserts and updates because these don't have performance issues and also because a lot of these are hidden and accessible due to abstractions from the CMS. –  Tiddo Sep 23 '13 at 14:30
1  
I just realized I misread you're answer. I thought you suggested to use views on the old schema. Using views on a new schema makes a lot of sense! Thanks! –  Tiddo Sep 23 '13 at 14:34
    
BTW, this is one of the core uses of views, to maintain app compatibility when db schemas change, IMO. –  Chris Travers Sep 24 '13 at 6:32

Your Answer

 
discard

By posting your answer, you agree to the privacy policy and terms of service.

Not the answer you're looking for? Browse other questions tagged or ask your own question.