GitLab.com stability post-GCP migration
Initial draft:
Andrew from GitLab here. I work in the Infrastructure Group: we’re responsible for running GitLab.com.
Thanks for your comments, we really appreciate the thanks massively and we will continue to work hard on improving the availability and stability of the platform. Our goal is currently to achieve 99.95% availability on GitLab.com.
According to Pingdom, over the last year our availability has been 99.81%, although this includes the large (multi-hour) maintenance window on 11 > August 2018 when we migrated GitLab.com from Microsoft Azure cloud to Google Cloud Platform (GCP).
Since the migration, our availability has improved greatly (caveat: we obviously have much less data than from Azure).
Using the data publicly available from Pingdom, here are some interesting stats:
- Mean-time between outages in Azure (September 2017 through August 11 2018): 1.3 days
- Mean-time between outages in GCP (September 2017 through August 12 2018): 7.3 days (if you ignore some problems we experienced on the first day after the migration, this rises to 12 days!)
Since, migrating to GCP, our overall availability, according to Pingdom, has risen to 99.92%. Again, if you exclude the few hiccups that we experienced on the Monday immediately following the migration, this rises to 99.97%.
There are multiple reasons for this improvement. We chose Google Cloud Platform because we believe that they offer the most reliable cloud platform for our workload, particularly as we move towards running GitLab.com in Kubernetes. It is worth pointing out that we also used the migration as an opportunity to improve our infrastructure, simplify some components and otherwise make things more stable and more observable. Finally, we’ve also been focusing on building the infrastructure team up, having hired many new team members over the past few months. This means that the team has been better able to balance the job of running GitLab.com with making it more stable.
Oh, and we’re still hiring! If you’re interested, check out https://about.gitlab.com/jobs/apply/ and select the Infrastructure Team.