In 2016, Microsoft acquired LinkedIn. Three years later, the “Blueshift” project was to ensure that the social network would make the move to Azure. Now, at least, the migration to Microsoft’s own Linux distribution Azure Linux is complete.
Site Reliability Engineer (SRE) at LinkedIn Ievgen Priadka states that almost all of the company’s servers, virtual machines and containers are now running on Azure Linux. The main reason for the switch was the fact that Red Hat’s CentOS 7 had an end-of-line date of June 2024. Diversifying the Linux distributions used was undesirable because it would have increased friction between teams.
From dated to progressive
The LinkedIn team was mostly satisfied with CentOS 7. Still, there were some problems that a replacement Linux distro shouldn’t have. For example, the user-space of Red Hat’s OS was outdated and not set up for modern apps. Also, bootstrapping an OS image was slow and updating security updates needed to be faster and automated. Vendors had also already partially walked away from CentOS 7, while the future of the free RHEL alternative looked uncertain.
LinkedIn’s new choice correctly focused on cloud adoption. Formerly called CBL-Mariner, this distribution has been known under the Azure Linux name since March. It is designed for Azure cloud infrastructure and edge services, as Microsoft points out on its GitHub page.
Challenges
Priadka’s explanation is comprehensive and technical; it effectively reads like a how-to guide for potential CentOS migrants toward Azure Linux. For example, the LinkedIn-SRE shows flowcharts that would map out a sensible migration path for many an organization. Other lessons in the piece include the earlier migration of stateful applications, which were easier to transfer, and discovering potential pain points early. For example, it took some time to choose the right file system because Apache Hadoop was not too friendly with XFS.
Extensive pilot programs prevented unforeseen problems in production. Developer VMs were available to users early thanks to the Productivity Engineering team, while other departments made significant preparations to prepare the security stack, critical packages and configuration modules. It was overwhelmingly a thought-out, patient process that anticipated CentOS 7’s end-of-life date. The question is whether other organizations have a similar luxury, especially if there is a staff shortage.
Deadline
The LinkedIn example reflects how, as an organization, one may prefer to modernize at a given time, but also how reality can bring this about earlier than one would have liked. Even at a Microsoft subsidiary, this decision to migrate started with a push factor: the end of CentOS 7 as a supported OS. Most other companies will depend on the goodwill and services of a party like SUSE, whose CEO Dirk-Peter van Leeuwen we spoke to recently. Indeed, he touched on the elephant in the room of the LinkedIn story: no one wants to be forced to reschedule their own roadmap due to external pressure.
Continue reading here: SUSE CEO: “If you want secure software, it has to be open source”