Thursday, 14 November 2013

Slides of My Talk at MySQL Percona Live London 2013

For those asking me about my talk at the Percona Live London Conference 2013, I'm making it available as PDF. Please feel free to contact me about it directly if you like.

Thank you all for showing up for the talk and for asking me so many questions. I was utterly pleased to answer all of your questions.

3 comments:

  1. Very nice Luis, thanks.
    I missed your presentation, so few questions here.
    1. Which File system do you use for data, backups partitions?
    2. Why RAID 5 and nor RAID 10 for data, backups partitions?
    3. What do you thins about LVM backup vs Extrabackup?

    I like COLLECTD + GRAPHITE + GDASH . looks absolutely gorgeous, gona give it a try.
    Do you guys use FK? Does it hurt the performance? I like the idea of using only PK on master and non PK indexes on slaves!

    Overall, great.

    ReplyDelete
    Replies
    1. 1. My default preferred filesystem is XFS. There will be a post about this soon enough.

      2. RAID5 offers me more space for the same amount of disks. As writes to the data partition should only happen in the background and never drive query time, I deem it as a good trade-off between more disk-space for a bit of I/O performance.

      3. I don't use LVM on my setup. I am not particularly found of the idea of "snapshot" as "hot-backup", because it involves having an intermediary database server somewhere where you can start a MySQL instance on your just-backed-up data so you can run the recovery procedure on it before you can actually restore from it. Despite the fact that XtraBackup does exactly the same procedure during it's preparation stage, I deem it a lot less error-prone than having to start a complete MySQL instance on your data to trigger a recovery.

      Delete
    2. I like LVM and use mylvmbackup script which is actually does recovery before zipping up. You can rsync snapshot to the remote box, in that case recovery is skipped. XFS is awesome. We moved away from ZFS as it couldn't keep up with our load of 6-8K QPS selects/writes 60%/40% .
      Waiting for more XFS stuff coming up!

      Delete