It’s time for the weekly roundup of news for Percona Server and XtraBackup. To follow up on my note from last week, we’re still checking and validating the download stats. It appears that Percona Server has been downloaded well over 100,000 times, and Percona XtraBackup has been downloaded over 200,000 times. But when we changed the script to separate out the downloads for our 5.5 release series, the overall number came back slightly decreased, which makes no sense. We haven’t had the chance to vet the analysis script yet, but we should be able to do that shortly.

For Percona Server,

  • We worked on Solaris builds for the server.
  • Peter proposed a feature for the server to enable more efficient incremental backups with XtraBackup.
  • Peter, Vadim, and Nickolay continued investigating alternative methods of implementing flush in Percona Server. (I am involved, but not as deeply.) The task is complex and there are a great many scenarios to consider. For example, there is fast storage and slow storage, there is the case when the data is larger than the buffer pool and when the data is smaller than the buffer pool (which is actually worse in some ways), there is flush list flushing and LRU flushing, and on and on. We have been benchmarking various types of scenarios for modified neighbor flushing, for example, to investigate how the oldest LSN is advanced in different cases. We still have nothing conclusive, but we have learned some things that gave us ideas about how we could make the flushing algorithm self-tuning in more cases. We will keep you posted.

In XtraBackup news,

  • Stewart Smith of the Drizzle project contributed a number of improvements to XtraBackup so that it can be a first-class backup tool for Drizzle. Thanks, Stewart!
  • Continuing with the theme of community contributions, are also delighted that Lachlan Mulcahy is working on software to manage backups with XtraBackup. It is called XtraBackup Manager and is hosted on Google Code.
  • We finished preparing XtraBackup to be used with Percona Server 5.5.
  • We worked on Solaris builds for XtraBackup.
  • We started from scratch and wrote documentation for innobackupex, removing mention of obsolete features and documenting command-line options that were not mentioned in the –help output previously. The docs will be available now with –help, via perldoc, as a man page, and on our documentation wiki.

In an unrelated note, we have a new home for the documentation of our InnoDB data recovery tools. The tools moved from Google Code to Launchpad a couple of years ago, but the documentation never got updated, and became stale and inaccurate. It is now hosted on our documentation wiki along with all of our other software documentation.