Scheduled maintenance Monday, April 11, 1am GMT up to one hour

By on April 8, 2011

We recently began to overhaul Bitbucket’s events system, which powers our dashboard page. We got caught up in all the NoSQL hype a while ago and began to store and retrieve events using Redis for the sake of speed and performance. Our Redis instance has grown to over 10GB, which is pretty big for a in-memory database. While Redis isn’t causing us any particular problems at the moment we’ve been looking at alternatives. We found that after some simple PostgreSQL configuration and query optimizations pulling events from our database is even faster than Redis.

Starting Monday, April 11st, 01:00 GMT we’re going to restart our PostgreSQL servers to pick up configuration changes to support our events system overhaul.

In the coming weeks we’ll deploy our revamped events system, which will display more events on your dashboard and allow you to go back further in time.  We expect downtime to be minimal, but we’re also using this opportunity to perform some other less interesting maintenance, such as package upgrades. The entire maintenance window will will last no longer than one hour.

Thanks for your patience as we work to increase Bitbucket’s performance and reliability.

10 Comments

  • Posted April 8, 2011 at 4:59 pm | Permalink

    Would you be willing to share some more technical details? We’re thinking of getting into the NoSQL hype so I’d like to know what we’re up against.

  • Posted April 8, 2011 at 4:59 pm | Permalink

    Would you be willing to share some more technical details? We’re thinking of getting into the NoSQL hype so I’d like to know what we’re up against.

  • Posted April 8, 2011 at 4:59 pm | Permalink

    Would you be willing to share some more technical details? We’re thinking of getting into the NoSQL hype so I’d like to know what we’re up against.

    • Charles McLaughlin
      Posted April 9, 2011 at 12:54 am | Permalink

      Petar,

      Redis is very cool and I wouldn’t advise against it. Our lead developer Jesper started using it as part of a custom load balancer. It worked well for that purpose because speed was really important. We switched to HAProxy for load balancing a few months ago, but keep using Redis to store our events. Currently our Redis instance is over 10GB and we want to store even more data. To be safe we dump it to disk every five minutes. Writing that much data to disk thrashes IO, which is the primary downside for us. We could work around that, but there’s really no need for us to choose anything other than a standard database at this point in time.

      Thanks for using Bitbucket!

      Charles

  • Posted April 8, 2011 at 9:13 pm | Permalink

    The need to upgrade your server(s) is the presage to success.

  • Posted April 8, 2011 at 9:13 pm | Permalink

    The need to upgrade your server(s) is the presage to success.

  • Posted April 8, 2011 at 9:13 pm | Permalink

    The need to upgrade your server(s) is the presage to success.

  • Shrawan Patel
    Posted June 8, 2011 at 11:24 am | Permalink

    This post excellently highlights what the author is trying to communicate. Nonetheless, the article has been framed excellently well and all credits to the author. For more information on how to load balance your web servers, please visit ..nhttp://serverloadbalancing.biz/wordpressbiz/, nhttp://serverloadbalancing.info/wordpressinfo/

  • Posted April 6, 2012 at 7:37 am | Permalink

    Good luck to you on improving the system!
    Upgrading your servers sounds like a great idea to optimize your work.

    Crossword puzzle maker http://thecrosswordpuzzlemaker.net/

  • Lidstromso
    Posted July 23, 2012 at 7:44 pm | Permalink

    The helm do not panic, the boat’s safe.

    White rice is delicious, grains difficult
    Tian kinds.

     

    Foamposites 2012

    Foamposites For Cheap

    http://www.foamposites2012online.com