When I started with MySQL 3.22 I would start running MySQL from early beta (if not alpha) and update MySQL the same date as release would hit the web. Since that time I matured and so did MySQL ecosystem. MySQL is powering a lot more demanding and business critical applications now than 12 years ago and it also a lot more complicated product with more things which can go wrong. I see a lot fewer customers running their systems on non GA software and a lot more customers running MySQL few major version behind the most current one – if you’re not running the most demanding or quickly changing application it well might be good enough.
Many users select rather traditional/prudent approach these days and will only start considering upgrade to next major MySQL version 6+ months after release. MySQL 5.5 is out almost a year ago and I would still estimate more people running on MySQL or Percona Server 5.1 than 5.5 Facebook also sticks with MySQL 5.1 even though their highly patched and improved version.
MySQL 5.0 which went to GA in October 2005 is in minority these days though it is not exceptional to see it with number of people jumping from version 5.0 straight to 5.5. This means 6 years since GA release date – quite respected.
The cases when you see old MySQL versions are used is typically when application is not overly demanding so performance improvements in recent releases is not critical, older applications which have not been actively developed recently are also more likely to run on old MySQL release which is understandable as upgrade requires extra QA effort might be application changes which might not be practical for old applications. For the same reasons we often see old MySQL versions not being upgraded to the most recent version in series. It is frequent to see something like MySQL 5.0.44 which went in as release in some of distributions.
As many people use MySQL releases which come with their distributions and some distributions would not offer MySQL upgrade to the next major version out of the box it is not a big surprise people running older Linux distributions are more frequently to stay on old MySQL releases.
Now where I’m looking for your feedback is understanding what kind of Version and Platform support do you think is adequate for MySQL ? Oracle seems to have extended support for MySQL versions rather significantly with Extended support going 8 years after release date. This does not seems to apply to MySQL 5.0 though which is described as EOL in the manual. even though it was released just about 6 years ago. With supported platforms it gets a bit more confusing. Per this table RHEL4 is Supported all the way to MySQL 5.5 yet it does not seems to be available from download pages. It looks like that page is outdated in more places. Support for Itanium was announced to be stopped yet that page still mentions it.
Do you think such policies are reasonable ? Do we need longer support cycle or shorter would make sense ? How long older platforms need to be supported ? I think for example it makes sense to support RHEL4 with MySQL 5.1 but support with MySQL 5.5 is not needed as people doing major MySQL upgrade usually would have an option to upgrade operating system as well
Or is MySQL Upgrade to new version while sticking to very old operating system is important for you ? Finally I’d like to know What is the Oldest Platform/Operating system are you running MySQL on now ? Do we have any RHEL4 still in production ? RHEL3 ?
I need your feedback so we can understand what policies would make sense for our builds and MySQL Support offerings. Supporting many platforms and versions for long time takes resources which alternatively could be used to do various cool stuff with Percona Server, Percona Xtrabackup and other Open Source software Percona develops.
I would appreciate your feedback on this matter either as comments to this blog posts or as direct email to my initials @percona.com