1. Maatkit almost makes me want to use Perl. But this post is about table checksums, not Perl. Statement based replication it possible to build simple and efficient tools for computing table checksums on MySQL deployments with many slaves. Maatkit does just that with mk-table-checksum. MONyog might do it as well.

    This works because the same SQL statement is run on the master and all of the slaves using the same view of the data. There is no need to stop replication on the slaves at the right point in time. However, this only works when statement based replication is used. Row based replication won't fix all problems that lead to inconsistent data. And even if you think it will, your diligence requires you to check to confirm your assumption.

    Falcon doesn't support statement based replication.
    1

    View comments

  2. This just arrived in my inbox. I expect a quiet debate, as only one participant is listed. The content might be interesting, but I can't stand webinars. Where is the white paper?


    EnterpriseDB, a Dr. Dobb's Journal partner, invites you to this complimentary Webinar.
    Event: The Great Debate: PostgreSQL versus MySQL
    Date: Tuesday, June 24, 2008
    Time: 11 am PT / 2 pm ET
    Duration: 60 Minutes
    Register now!

    Overview:
    For years, the common industry perception has been that MySQL is faster and easier to use than PostgreSQL. PostgreSQL is perceived as more powerful, more focused on data integrity, and stricter at complying with SQL specifications, but correspondingly slower and more complicated to use.

    Like many perceptions formed in the past, these things aren't as true with the current generation of releases as they used to be.

    DBAs, developers, and IT managers and decision-makers will benefit from this hour-long presentation about the pros and cons of using PostgreSQL or MySQL, which will include a discussion about the ongoing trend towards using open source in the enterprise. Questions will be answered upon completion of this presentation.

    Speaker:
    Jim Mlodgenski, VP Worldwide Technical Services, EnterpriseDB Corporation
    Jim is one of EnterpriseDB's first employees and joined the company in May 2005. As vice president, worldwide technical services, he is responsible for EnterpriseDB's technical pre-sales, professional services, and training. Prior to joining EnterpriseDB, Jim was a partner and architect at Fusion Technologies, a technology services company founded by EnterpriseDB's chief architect, Denis Lussier. For nearly a decade, Jim developed early designs and concepts for Fusion's consulting projects and specialized in Oracle application development, Web development, and open source. Jim received a BS degree in Physics from Rensselaer Polytechnic Institute.


    0

    Add a comment

  3. Hidden enterprise.
    Waste 2 minutes downloading.
    How many of us test it?

    Sorry for the bad haiku. But why is the source for Enterprise MySQL releases locked up behind support.mysql.com? It takes me an extra 2 minutes to download and time is money. If I were an Enterprise customer, and maybe I am, I would sleep much better knowing that as many people as possible were using and testing it. You might even get the community to do more QA for MySQL for free!
    6

    View comments

  4. Notes for those trying to setup a Dell laptop with Ubuntu Hardy Heron. I have a Dell Vostro. Pardon the interruption from the normal MySQL programming. I guess I can make this related to databases by comparing the ease of installing and setting up MySQL including replication with the difficulty of doing the same for some of the commercial products that came before it.
    • the screen was off-white on resume after suspend or hibernate. This is fixed by disabling lock on suspend and hibernate.
    • wireless with a Dell Wireless 1490 chip didn't work by default. This is fixed by installing b43-fwcutter.
    • Load_Cycle_Count was increasing too quickly on my hard drive. This is fixed by enabling laptop mode (set ENABLE_LAPTOP_MODE=true in /etc/default/acpi-support).
    The last problem has a huge amount of information and FUD on the web. The basic problem appears to be that the disk drive head is parked too frequently on modern laptops for many OS distributions (both free and proprietary). By too frequently, I mean that the number of head parking events will exceed manufacturer's limits in less than 1 year. Ubuntu has been blamed for this, but the problem is not limited to Ubuntu.

    For my install, enabling laptop mode fixes the problem. By enable, I mean that I set ENABLE_LAPTOP_MODE=true in /etc/default/acpi-support. Laptop mode was already enabled in /proc/sys/vm/laptop_mode. With this change, Load_Cycle_Count as measured by hdparm does not increase when the laptop runs on either battery or AC. With battery, the APM level is set to 1 which enables spindown and parking. With AC, the APM level is 254 which disables head parking and spindown. I still need to check battery temperature with the APM level at 254. APM level is the value displayed by hdparm for Advanced Power Management and the value used for hdparm -B. Changing the APM level manually without using ENABLE_LAPTOP_MODE made the problem worse.

    Alas, this was a lot of work. It will be nice when this is handled by the install tool.

    This post has a good summary.
    0

    Add a comment

  5. I have created a project at Launchpad to publish my branch of MySQL 5.0. Right now it is MySQL 5.0.37 and the Google patch. It is likely to be limited to tracking the Google patch for some time.

    There are other interesting MySQL branches at Launchpad. My favorite is the Wikimedia edition.
    4

    View comments

  6. The PostgreSQL core team has announced their intention to integrate replication into the PostgreSQL core in the next two releases (8.4, 8.5). There isn't a design yet, but the announcement mentions:
    • asynchronous replication
    • transaction log shipping
    • slaves will be cold standby servers in 8.4 and can be queried in 8.5
    There was a lot of discussion that followed the announcement with a variety of opinions about what is needed and what is useful. Features that were debated include synchronous versus asynchronous replication and the value of supporting read-only slaves versus slaves that cannot be queried when replaying transaction logs. I hope that async replication and read-only slaves are supported.
    0

    Add a comment

Loading