For much too long I’ve been working on getting the pool system running under kubernetes. The beta site has been running under k8s for a while. I have a couple more tweaks to do and then I plan on moving the production site (www.ntppool.org / www.pool.ntp.org ) as well.
The old setup was a “traditional” setup with puppet configuration management on dedicated servers. This has been great for a long time, but certain upgrades and tests were hard to do. It also made it more tedious than necessary to split up the application in smaller pieces when that made sense.
I forget what exactly, but there was something that I needed to upgrade way back in 2016 that proved hard to do separately on the beta and production systems (they share the underlying hardware) and it put me on the path to tinker with this. I had a while in 2017 where I wasn’t able to do much work and when I started again, puttering with kubernetes was about the right level of work for me. The (very small) kubernetes cluster has been running for about 9 months with what seems like a reasonable problems-to-benefits ratio.
After the production site is moved some servers will be freed up to be added to the kubernetes cluster.
I’m pre-announcing this in case it’s causing trouble so I don’t have to type up the explanation then and also because I hope having this completed will make it easier for me to find time for some of the more worthwhile tweaks and changes that have been requested and are in the backlog.