I've been using Amazon RDS instances for Postgres databses for a while, but I haven't had to do a lot of upgrading on large databases until recently. I hit an issue where one database was hundreds of gigs large. The upgrade never completes because the instance is oom killed. I've been reading a lot on the AWS website about common practices for upgrading databases major versions , but I haven't seen anything about large databases oom killing the process and how to avoid.

I believe at least one of our databases on this instance have the potential to eventually get up to 500 gigs or so. Obviously scaling instances vertically to that much memory isn't a great solution (and probably doesn't exist). Also, scaling horizontally JUST for memory to upgrade doesn't seem to be a good approach. What is the normal way to handle this?

  • FWIW, I'm surprised that the size of the database comes into play with a database system version upgrade. I think that's hardly the case with SQL Server. How have you determined that the size of your database is relevant to the issue?
    – J.D.
    Feb 7 at 19:03
  • I guess it's possible that it's not actually the cause. But, I have multiple RDS instances of identical configuration. I've upgraded all of them through the same versions, and the only one that fails is the one that has a database that's very large. And using monitoring tools I can see that the upgrade tries, until the instance is oom killed. There's no differences between all of the instances except for the size of that one database. I've also ruled out several other things that I thought it was initially. Feb 7 at 19:06
  • I cannot imagine an upgrade procedure that needs lots of RAM. Since you cannot directly see what is going on, you might have to ask Amazon. Feb 8 at 16:36


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.