IBM buys the biggest open source company in the world, and GitHub had a long outage.

IBM buys Red Hat

I did not see this coming, but the largest software acquisition ever got announced last week. IBM is buying Red Hat.

Now, I feel like I’ve been pretty positive about acquisitions lately but in all honesty I can’t see anything good coming out of this. And it also means I have to reconsider my opinion about RedHat buying CoreOS earlier in the year. I suspect the people there feel a bit like MySQL did after Sun got bought by Oracle.

The press release focuses a lot on the Hybrid Cloud part of IBM, which seems to be their answer to not being able to compete with the real cloud providers. If this sounds like I don’t see a future for Hybrid Cloud, you’re understanding correctly. It works perfectly fine as a stopgap, but eventually it doesn’t actually do anything about the problem that cloud is supposed to solve¹.

The other reason for buying the company is likely to be around Red Hat’s strength in the Kubernetes space². With both Openshift and the aforementioned CoreOS offerings there is indeed a case to be made there. Except, as pointed out by Mark Shuttleworth, enterprises don’t really care as much about the shrink-wrapped solutions anymore³.

Where it comes to the possible reasoning for acquisitions like this however, I always enjoy reading Stratechery and Ben Thompson certainly doesn’t disappoint with his reasoning behind this one.

GitHub outage

Speaking of acquisitions, Microsoft’s acquisition of GitHub was finalized last week and somehow the world managed to not burn down⁴.

More interesting is the major outage of 24 hours and 11 minutes⁵ that GitHub experienced a couple days earlier. As this isn’t exactly a common occurrence it’s worth having a look at what happened. Luckily, GitHub posted a great post-incident analysis. I recommend you read that, but I’ll summarize it anyway.

During some routine maintenance to replace some failing equipment an outage of 43 seconds happened between their 2 main data centers. This in result caused the two zones to have separate master databases that were both being written to. When the connection was then restored, the databases flagged they couldn’t reconciliate this and alarms started going off.

Within GitHub the response was very fast, and they made the decision that data integrity was more important than offering full functionality and so disabled parts of the site while working on ensuring all the data was reconciled. While this had a big impact, I do believe they made the right call⁶ and I always appreciate it when a company posts a detailed post-incident analysis.